fix(lint): check resp.Body.Close()'s error (errcheck)

Caught by CI, not locally -- this sandbox's golangci-lint crashes on any repo, a pre-existing environment issue unrelated to this code (documented in dispatch's own session history). gitea-mcp's actual CI run has a working linter and correctly flagged this.
feat(admin): local-only stdio MCP server to mint scoped service-account tokens (dispatch#67)
2026-09-05 11:27:22 +02:00 · 2026-09-05 11:25:41 +02:00 · 2026-09-05 08:01:07 +02:00 · 2026-08-25 17:25:23 +02:00 · 2026-08-25 15:29:40 +02:00 · 2026-08-25 12:35:52 +02:00
62 changed files with 1660 additions and 930 deletions
@@ -1,2 +0,0 @@
-read: .aider.conventions.md
-auto-commits: false
@@ -1,386 +0,0 @@
-# Agent context — Mathias workspace
-
-<!-- Canonical root context for all AI coding agents.
-     Lives at: ~/dev/.context/AGENT.md
-     Applies to every project under ~/dev/ unless overridden.
-     
-     Run `task context:sync` from ~/dev/ to regenerate harness-specific files.
-     Project-level context in .context/PROJECT.md layers on top of this. -->
-
-## Who I am
-
-I'm Mathias, a digital product manager and technology consultant based in Sweden.
-I build software, research emerging tech, and deliver consulting engagements
-for clients under NDA. I work across AI/ML, financial automation, web applications,
-and climate/sustainability tech.
-
-## How I work with agents
-
- I think like a product manager — I care about *why* before *how*
- I want agents to be opinionated and push back, not just execute blindly
- I prefer concise responses; skip ceremony and get to the point
- When I say "build this", I mean production-quality with tests, not a demo
- Ask me before making irreversible changes or adding heavy dependencies
- I work with confidential client data — never send it to cloud APIs unless I explicitly say it's OK
-
-## Behavior rules
-
-These rules apply to every task across every project, regardless of harness.
-
-0. **Pre-task ritual — before ANY implementation (non-negotiable).** Run this before writing a single line:
-   - **Query the brain** (`brain_query`) for the domain + symptom. If the result changes your approach, surface it before acting. 5 seconds beats 5 hours.
-   - **Load the relevant skill** — see trigger table in *Engineering Skills* below.
-   - **Write the failing test first.** Name the test before the function. If the target is untestable (e.g. `main()` wiring), extract the logic into a testable function first. No implementation without a red test.
-   - **State the observable success criterion** — what specific behavior, output, or passing test proves this is done?
-
-   **TDD is non-negotiable.** "Tests pass" is not proof of correctness — only proof the tests ran. Write tests that would catch the bug before writing code that fixes it.
-
-1. **No assumptions.** Don't hide confusion — surface it. Surface tradeoffs explicitly.
-   Think before coding; if the problem is unclear, ask or state assumptions before acting.
-2. **Minimum viable code.** Solve with the smallest change that works. Nothing
-   speculative, no "while we're here" cleanups, no premature abstractions. Simplicity first.
-3. **Surgical changes.** Touch only what the task requires. Leave unrelated code,
-   files, and formatting alone. Diffs should be small and reviewable.
-4. **Goal-driven execution.** Define clear success criteria up front for every task.
-   Loop — implement, verify, refine — until those criteria are met. Don't claim
-   completion without evidence (tests pass, command output, observed behavior).
-5. **Trunk-Based Development — commit directly to main.** Every commit is one
-   logical change (one tool, one fix, one test) with passing tests. Main is always
-   deployable. Never create long-lived feature branches.
-
-   **Exception — parallel agents on same repo:** If another agent is known to be
-   actively working on the same repo simultaneously, create a short-lived branch
-   (`agent/<description>`), finish the task, and merge to main within the same
-   session. Do not leave agent branches open between sessions.
-
-   **Exception — external contributor or client four-eyes requirement:** Use
-   PR flow only when a human reviewer outside the project is required. Document
-   the reason in PROJECT.md.
-
-6. **Close the loop — every substantive task ends with the same ritual.** Shipping
-   the code is not the end of the task; capturing it is. Run this unprompted:
-   - **Tag + bump SemVer** on the change (annotated tag; minor for a feature or
-     new/changed ADR, patch for a fix; docs in the same commit). Check the repo's
-     actual last tag — stated versions in docs drift stale.
-   - **Push** main and the tag (CI is the gate).
-   - **Persist generalizable learnings to the brain** (`brain_write`, wing/hall) —
-     the reusable patterns and the footguns that would bite anyone again, never
-     project status. See *Knowledge base — when to write* below.
-   - **File discovered-but-deferred work as tracker issues** on the project's own
-     repo — token-budget gaps, recorded ADR limitations, v2 follow-ups. Don't let
-     "out of scope, recorded" rot in a commit message; make it a ticket with a
-     source pointer.
-   - Surface the brain entries and issue numbers in the closing summary so the
-     trail is auditable.
-
-## Default stack
-
-| Layer | Default | Fallback | Last resort |
-|-------|---------|----------|-------------|
-| Language | Go | Python | TypeScript, Java, C |
-| UI | HTMX + Templ | Server-rendered HTML | React (only if SPA is justified) |
-| Build | Task (taskfile.dev) | Make | — |
-| Containers | Docker Compose (dev), k3s (prod) | — | — |
-| DB | PostgreSQL + sqlc | SQLite | — |
-| Search | pgvector (vector), BM25 | Qdrant (when >1M vectors or hybrid retrieval) | — |
-| Logging | slog (structured) | — | — |
-| Testing | Table-driven, testify | — | — |
-| Agents (Go) | google.golang.org/adk + pkg/litellm adapter | — | — |
-
-Exploratory: Rust, Zig — I'll tell you when I want these.
-
-## Code conventions
-
- **Go style**: golines, gofumpt, golangci-lint
- **Errors**: `fmt.Errorf("operation: %w", err)` — never naked, never log-and-return
- **Naming**: stdlib conventions, no stuttering
- **Architecture**: prefer stdlib over frameworks, constructor injection, env-var config parsed into typed structs
- **Git**: conventional commits (`feat:`, `fix:`, `chore:`), commit directly to main,
-  one logical change per commit, CI is the quality gate
- **Never**: long-lived feature branches, PRs for solo work, direct push without
-  passing `task check` locally first
- **Security**: no secrets in code, govulncheck before adding deps, SOPS for encrypted config
- **Dependencies**: prefer stdlib. testify, slog, templ, sqlc, google.golang.org/adk (agent projects only) are pre-approved; anything else needs justification in the commit message
-
-## Secret handling (every harness, every command)
-
-Tool output is persisted: terminal → `~/.claude/projects` transcripts →
-claudewatcher → brain/wiki → gitea history. A secret printed once is
-searchable forever, and clearing it means rotating the key. So:
-
-1. **Never print, echo, log, or transform a secret to inspect it.** No
-   `base64`/`xxd`/`cat` of a key, and never pipe a secret through a transform
-   to defeat `op run`'s output masking (it masks raw values; base64 hides them
-   from the mask — that exact trick leaked a key on 2026-06-11).
-2. **Secrets stay in the subprocess.** Reference them only as env vars consumed
-   *inside* `op run --env-file ~/.op-env -- <cmd>`. Never place a literal secret
-   in a command's argv (it lands in the tool call and the transcript).
-3. **Existence check without revealing the value:** `[ -n "$X" ] && echo set` —
-   never `${X:-...}` (returns the value when set) and never echo a substring of it.
-4. **Cross-host secrets:** run the secret-consuming command on the host that has
-   the secret; do not forward a raw key over ssh argv/stdout.
-5. If a secret does leak into output, say so immediately and flag it for rotation —
-   don't bury it.
-
-## Infrastructure
-
-Three machines on Tailscale:
-
-| Machine | Role | Key specs |
-|---------|------|-----------|
-| koala | GPU inference, heavy compute | RTX 5070, runs k3s + llama-swap + shared postgres18/pgvector |
-| iguana | Services, builds | M2 Ultra Mac |
-| flamingo | Daily driver, edge | Mac mini, ~/dev is here |
-
- **Model routing**: LiteLLM in front of llama-swap (local) + cloud APIs (when permitted)
- **Orchestration**: k3s cluster across all three machines
- **Networking**: Tailscale mesh
-
-## Project landscape
-
-All development repos live at `~/dev/` (softlink from `~/Documents/local-dev/`).
-
-Organized in thematic folders:
-
-| Folder | Focus | Count |
-|--------|-------|-------|
-| `GO/` | Go web frameworks, API integrations, learning projects | ~10 |
-| `AI/` | ML research, AI frameworks (FinRL, DSPy, crawl4ai) | ~6 |
-| `AGENTS/` | Autonomous agents, coding agents, MCP servers, infra | ~15 |
-| `QKX/` | Invoice processing, financial automation, payment systems | ~13 |
-| `XT/` | Climate data, sustainability (Klimatkollen, Garbo) | ~2 |
-
-See `~/dev/PROJECT_SUMMARY.md` for detailed descriptions of each project.
-
-### Key active projects
-
- **super-koala** (`AGENTS/`) — multi-component agent stack with LangGraph, DSPy, MCP
- **azure-tiger** (`QKX/`) — invoice extraction → ISO 20022 payment instructions
- **gocrwl** (`AGENTS/`) — Go web crawler with containerized deployment
- **koala-ai-stack** (`AGENTS/`) — local AI server infrastructure management
- **klimatkollen** (`XT/`) — Swedish municipal climate data platform
-
-## Knowledge base — actively use it
-
-A persistent brain (BM25 search + LLM-synthesised Q&A) survives across sessions,
-hosts, and harnesses. It holds 100+ hard-won entries: infra incident postmortems,
-Go pitfalls, framework gotchas, design principles, ADRs. **It is not optional
-reference material — query it actively, not just when explicitly told.**
-
-### When to query (treat as a reflex)
-
- **Before** starting a non-trivial task — search for prior art with the symptom
-  AND the system component ("how did we solve X in Y?"). 5 seconds beats 5 hours.
- **When debugging** — search for the error string, the stack frame, the affected
-  service. Past you may have already paid this tax.
- **Before adopting** a pattern, library, framework, or model name — check if it
-  was tried and rejected, or what the integration footguns are.
- **When making architectural decisions** — search for the domain + "ADR" or
-  "decision" to find prior reasoning before re-deriving it.
- **When a recommendation feels novel** — challenge yourself: "has this been
-  documented?" The brain often has it.
-
-### When to write
-
-After you discover something that **future-you would forget** and that **isn't
-recoverable from the code, git log, or PR description alone**:
-
- Bugs whose root cause is non-obvious and generalisable beyond this project.
- Framework / library / model-name quirks that bit you and would bite anyone.
- Design principles validated under fire (e.g. "every `_get` needs a `_list`").
- Postmortems for incidents: what broke, why, how diagnosed, what to do next time.
-
-DON'T write project status, sprint progress, PR summaries, or "what I did this
-session" — those rot fast and the originals are in git/gitea anyway. Brain
-entries that age well are about *why*, *how to avoid*, and *what to do when*.
-
-### How to access (per harness)
-
-| Harness | Query | Write |
-|---------|-------|-------|
-| **Claude Code, Claude Desktop** | `brain_query` (BM25), `brain_answer` (LLM-synth + sources) MCP tools | `brain_write` MCP tool |
-| **Crush, Pi, Antigravity, other MCP-capable** | same MCP server: `ingestion-brain` (via the `mcp__*_brain__*` namespace once authenticated) | same |
-| **Anything HTTP-only (curl, scripts)** | `POST https://brain-mcp.d-ma.be/query` with `{"query":"..."}` (auth via `BRAIN_MCP_TOKEN`) | `POST .../write` with `{"content":"...","filename":"..."}` |
-| **Browser / human inspection** | `https://git.d-ma.be/mathias/hyperguild` → `knowledge/` and `wiki/` markdown files |
-
- **Scoping**: defaults to `public` collection; client projects filter to `{client}` + `public`.
- **Routing**: brain_answer's LLM uses berget.ai as primary, iguana ollama as
-  fallback. Both are configurable in the `supervisor/ingestion-deployment.yaml`
-  on the koala k3s cluster; don't hardcode local-only model names into the
-  berget URL (see knowledge entry on namespace mismatches).
-
-### Quick reflex checks
-
-If you find yourself about to say any of these out loud, you owe yourself a brain query first:
-
- "I think the issue might be..."
- "Let me try X and see..."
- "I'll just write a script to..."
- "This is probably a new bug..."
- "Has anyone done this before?" — *yes, probably, go check.*
-
-## Client work rules
-
-When working on a project tagged with a client name:
-1. Never send code, data, or context to cloud APIs — use local models only
-2. Never reference other client projects or their data
-3. Keep all artifacts within the client's git org / directory
-4. Treat everything as confidential unless told otherwise
-
-## Harness-agnostic principles
-
-This context is designed to work with any AI coding tool:
- Claude Code, Cursor, Aider, Open WebUI, Charmbracelet Mods/Crush
- Pi Coding Agent, Mistral Vibe, Antigravity
- Any tool that accepts a system prompt or reads a markdown context file
-
-The canonical source is always `.context/AGENT.md` (root) and `.context/PROJECT.md` (per-project).
-Derived files are committed (see *How context propagates* below) so a `git pull` on any host yields full agent context with no setup.
-
-## How context propagates
-
-Canonical sources of truth:
- Universal: `~/dev/.context/AGENT.md` (this file)
- Project: `<repo>/.context/PROJECT.md` (per-repo)
-
-Derived files (committed, regenerated by `task context:sync`):
- `CLAUDE.md`, `AGENTS.md`, `.cursorrules`, `.aider.conventions.md`,
-  `.context/system-prompt.txt`
-
-Workflow:
-1. Edit a canonical file. Run `task context:sync`. Commit canonical and
-   derived together. Push.
-2. On any other host, `git pull` brings both. Claude Code (tree-walking)
-   uses `CLAUDE.md`; Crush / Pi / Antigravity (cwd-only) use `AGENTS.md`;
-   Cursor uses `.cursorrules`; Aider uses `.aider.conventions.md`.
-3. `task check` runs `context:sync` then asserts `git status --porcelain`
-   is empty over the derived files (catches both modified-tracked drift
-   and missing-untracked adapters). A drift fails the check with a
-   message telling you to stage the regenerated files.
-
-Behavior rules in this file and per-project rules in `PROJECT.md` apply
-unconditionally on every host, every harness.
-
-## Engineering Skills
-
-Shared engineering skills live in the **`mathias/skills`** repo (`git.d-ma.be/mathias/skills`). Clone it to `~/dev/skills/` and run `SKILLS_CHECKOUT_DIR="$PWD" bash install.sh` there to wire every skill into your harnesses (Claude Code, Crush, Antigravity, Mistral Vibe) as native, on-demand skills. (Use `install.sh`, not `task install` — the latter is currently broken, skills#7.) Load at task start — not "on demand" but on schedule, before writing code. Browse `~/dev/skills/SKILLS_INDEX.md` for the full list.
-
-**Skill trigger table — load before starting, not after getting stuck:**
-
-| Task type | Load |
-|-----------|------|
-| Any feature or bug fix | `tdd` |
-| Refactor or design | `clean-code` or `solid` |
-| Debug | `problem-analysis` |
-| Review code or PRs | `code-review` |
-| Frame a problem before coding | `problem-analysis` |
-
---
-
-# Project context
-
-<!-- Canonical project context. Edit this, run `task context:sync`.
-     Root agent context from ~/dev/.context/AGENT.md is automatically
-     prepended for harnesses that don't walk the directory tree. -->
-
-## Identity
-
- **Name**: gitea-mcp
- **Owner**: Mathias
- **Client**: personal
- **Repo**: https://git.d-ma.be/mathias/gitea-mcp
- **Status**: active
-
-## Stack
-
- **Primary language**: Go
- **UI layer**: HTMX + Templ (when applicable)
- **Fallback languages**: Python, TypeScript (justify in PR if used)
- **Build**: Task (taskfile.dev), not Make
- **Containers**: Docker (compose for dev, k3s for deploy)
- **Target infra**: koala (GPU workloads), iguana (services), flamingo (edge)
-
-## Conventions
-
-### Code style
- Go: follow `golines`, `gofumpt`, `golangci-lint` with project config
- Tests: table-driven, in `_test.go` next to source, `testify` for assertions
- Errors: wrap with `fmt.Errorf("operation: %w", err)`, no naked returns
- Naming: stdlib conventions, no stuttering (`http.Client` not `http.HTTPClient`)
-
-### Architecture preferences
- Prefer standard library over frameworks (net/http over gin/echo)
- Dependency injection via constructor functions, not containers
- Configuration via environment variables, parsed at startup into a typed struct
- Structured logging via `slog`
-
-### Git
- Conventional commits: `feat:`, `fix:`, `chore:`, `docs:`, `refactor:`
- **Trunk-Based Development:** commit directly to main. One logical change per commit.
- Run `task check` locally before every push. CI is the quality gate, not branch protection.
- No feature branches, no PRs for solo/agent work.
- Exception: if a parallel agent session is active on this repo, use a short-lived
-  `agent/<description>` branch and merge within the same session.
-
-### Security
- No secrets in code, ever — use env vars or SOPS-encrypted files
- Client data never leaves local network unless explicitly cleared
- Dependencies: audit with `govulncheck` before adding
-
-## Knowledge base access
-
-This project can query the shared knowledge base via MCP or HTTP:
-
- **MCP endpoint**: `mcp://localhost:3100/knowledge`
- **HTTP fallback**: `http://localhost:3100/api/v1/search`
- **Scoping**: queries are filtered to collection `personal` + `public`
-
-## Behavior rules
-
-These rules apply to every task in this project, regardless of harness.
-
-1. **No assumptions.** Don't hide confusion — surface it. Surface tradeoffs explicitly.
-   Think before coding; if the problem is unclear, ask or state assumptions before acting.
-2. **Minimum viable code.** Solve with the smallest change that works. Nothing
-   speculative, no "while we're here" cleanups, no premature abstractions. Simplicity first.
-3. **Surgical changes.** Touch only what the task requires. Leave unrelated code,
-   files, and formatting alone. Diffs should be small and reviewable.
-4. **Goal-driven execution.** Define clear success criteria up front for every task.
-   Loop — implement, verify, refine — until those criteria are met. Don't claim
-   completion without evidence (tests pass, command output, observed behavior).
-
-## Agent instructions
-
-When acting as a coding agent on this project:
-
-1. Read this file and all `SKILL.md` files in `.skills/` before starting work
-2. Run `task check` before committing (lint + test + vet)
-3. If unsure about a convention, check `DECISIONS.md` or ask
-4. Never modify files outside the project root without explicit permission
-5. When adding a dependency, explain why in the commit message
-6. Commit directly to main. Run `task check` before every push. Never create
-   feature branches unless a parallel agent is simultaneously active on this repo.
-7. For client projects: never send code or context to cloud APIs — use local models via LiteLLM
-
-## Current state — v0.2.5 (2026-05-17)
-
-All v0.2 work is complete and deployed. No active sprint.
-
-### What shipped
-
-| Tag | PR | Tools / fixes |
-|-----|----|---------------|
-| v0.2.2 | #21 | repo_create, repo_update, repo_mirror_push |
-| v0.2.3 | #22 | repo_tree, repo_topics_update, file_read dir fix |
-| v0.2.4 | #23 | issue_get, release_create, repo_delete |
-| v0.2.5 | #26 | repo_update archived+template, create_project_from_template template_name, pr_files_diff loop fix |
-
-Current main: `e31fd3f`. CI green. Deployed via Flux.
-
-### Next up
-
-1. **`hyperguild new-project` v1** — primary next target.
-   See brain node `adr-new-project-gitea-first-github-mirror` for full flow spec.
-
-2. **Issue #19** — end-to-end mirror flow verification.
-   `repo_mirror_push` is implemented but the full flow (create repo → add push mirror → verify sync to GitHub) has not been tested manually. Do this before relying on it in production.
@@ -53,14 +53,30 @@ These rules apply to every task across every project, regardless of harness.
   logical change (one tool, one fix, one test) with passing tests. Main is always
   deployable. Never create long-lived feature branches.

-   **Exception — parallel agents on same repo:** If another agent is known to be
-   actively working on the same repo simultaneously, create a short-lived branch
-   (`agent/<description>`), finish the task, and merge to main within the same
-   session. Do not leave agent branches open between sessions.
+   **The PR trigger is RISK, not concurrency.** `risk/low` work commits straight to
+   main with CI as the gate — that is most work and is unchanged. `risk/medium` and
+   above opens a PR, even solo, even with nobody waiting to review it. This mirrors
+   the tiered merge model in `assessor-loop`
+   `method/transfer-to-a-regulated-organisation.md` §4.1, so the estate practises
+   what it sells instead of merging everything on green and claiming otherwise.

-   **Exception — external contributor or client four-eyes requirement:** Use
-   PR flow only when a human reviewer outside the project is required. Document
-   the reason in PROJECT.md.
+   **Concurrency is NOT a reason to branch.** Measured 2026-08-29 across three real
+   collisions between parallel sessions: two were not branching problems at all — a
+   generated artifact written to the same path by both (branching turns that into a
+   merge conflict on a 33k-line JSON), and uncommitted work in a shared checkout
+   (a per-session worktree fixes it) — and the third was an ordinary non-fast-forward
+   push resolved by rebase in seconds. Use a worktree per session on shared repos and
+   timestamp generated artifacts. Do not reach for a branch.
+
+   Branch+PR flow carries documented costs on solo repos: merging auto-closes gated
+   issues, review comments live on a thread separate from the issue's, and a
+   merged-and-deleted branch cannot be amended (brain:
+   `gitea-pr-merge-autoclose-and-review-comment-footguns`). Those are the price of
+   the `risk/medium`+ lane — paid deliberately, not incurred by accident.
+
+   **Exception — external contributor or client four-eyes requirement:** Use PR flow
+   when a human reviewer outside the project is required. Document the reason in
+   PROJECT.md.

 6. **Close the loop — every substantive task ends with the same ritual.** Shipping
   the code is not the end of the task; capturing it is. Run this unprompted:
@@ -78,6 +94,26 @@ These rules apply to every task across every project, regardless of harness.
   - Surface the brain entries and issue numbers in the closing summary so the
     trail is auditable.

+   **Chat is not a durable store. Never rely on Mathias remembering anything.**
+   Every gap, caveat, risk, open question and next step must land in one of four
+   places before the turn ends: **the brain**, **a repo** (code, docs, ADR),
+   **a tracker issue**, or **a plan file**. A finding stated only in conversation is
+   lost the moment the session closes — and this estate has repeatedly proved that a
+   remembered convention is an unreliable control.
+
+   This applies hardest to what is tempting to leave as a closing remark: "worth
+   knowing X isn't enforced yet", "this is a convention, not a gate", "we should
+   revisit Y". Those are exactly the items that need a locator. If it is worth
+   saying in a summary, it is worth writing where it can be found again without you.
+
+## CAD sprint patterns
+
+Two CAD sprint archetypes (defs: brain `wiki/homelab/decisions/continuous-agentic-development-cad-concept-2026-06-16`):
+- **Inception Sprint** — birth a new repo. Oath-gated close; tag only once every clause holds. Brain: `.../inception-sprint-and-oath`.
+- **Sweep Sprint** — clear an existing backlog. Parallel read-only research per issue → sequential one-issue-per-commit execution. Brain: `.../sweep-sprint-and-oath`.
+
+Shared: decisions that are yours, not the agent's, surface as `DECISION NEEDED` (question/options/recommendation/blocks) — ask live if you're present, else comment and wait, never act past it. Live-breaking regressions get `git revert`, never forward-fixed; host-network work skips even read-only research (solo, synchronous, you present). Rule 6's closing ritual only fires once every Oath clause genuinely holds — blocked = descoped-and-tracked, never marked done.
+
 ## Default stack

 | Layer | Default | Fallback | Last resort |
@@ -88,7 +124,7 @@ These rules apply to every task across every project, regardless of harness.
 | Containers | Docker Compose (dev), k3s (prod) | — | — |
 | DB | PostgreSQL + sqlc | SQLite | — |
 | Search | pgvector (vector), BM25 | Qdrant (when >1M vectors or hybrid retrieval) | — |
-| Logging | slog (structured) | — | — |
+| Logging | slog (structured) | stdlib `logging` w/ structured `extra=` (or structlog) | — |
 | Testing | Table-driven, testify | — | — |
 | Agents (Go) | google.golang.org/adk + pkg/litellm adapter | — | — |

@@ -97,7 +133,12 @@ Exploratory: Rust, Zig — I'll tell you when I want these.
 ## Code conventions

 - **Go style**: golines, gofumpt, golangci-lint
- **Errors**: `fmt.Errorf("operation: %w", err)` — never naked, never log-and-return
+- **Python style** (fallback language): ruff (format+lint, one tool), mypy --strict (non-negotiable,
+  matches Go's static typing discipline), pytest + pytest-cov (table-driven via
+  `@pytest.mark.parametrize`), uv (venv+deps+lock, one tool), pydantic-settings (typed env-var config
+  — same principle as Go's typed structs), src-layout + `pyproject.toml` only (no `setup.py`)
+- **Errors**: `fmt.Errorf("operation: %w", err)` — never naked, never log-and-return.
+  Python: `raise X from e` (exception chaining, same principle) — never bare `except`, never silent `pass`
 - **Naming**: stdlib conventions, no stuttering
 - **Architecture**: prefer stdlib over frameworks, constructor injection, env-var config parsed into typed structs
 - **Git**: conventional commits (`feat:`, `fix:`, `chore:`), commit directly to main,
@@ -135,10 +176,10 @@ Three machines on Tailscale:
 |---------|------|-----------|
 | koala | GPU inference, heavy compute | RTX 5070, runs k3s + llama-swap + shared postgres18/pgvector |
 | iguana | Services, builds | M2 Ultra Mac |
-| flamingo | Daily driver, edge | Mac mini, ~/dev is here |
+| flamingo | Daily driver, edge | Apple Silicon Mac, ~/dev is here |

 - **Model routing**: LiteLLM in front of llama-swap (local) + cloud APIs (when permitted)
- **Orchestration**: k3s cluster across all three machines
+- **Orchestration**: k3s on koala only — single control-plane node, no workers (verified 2026-08-30). iguana and flamingo are on the tailnet but are not cluster members
 - **Networking**: Tailscale mesh

 ## Project landscape
@@ -234,10 +275,11 @@ When working on a project tagged with a client name:

 ## Harness-agnostic principles

-This context is designed to work with any AI coding tool:
- Claude Code, Cursor, Aider, Open WebUI, Charmbracelet Mods/Crush
- Pi Coding Agent, Mistral Vibe, Antigravity
- Any tool that accepts a system prompt or reads a markdown context file
+Supported harnesses (2026-09-01) — Cursor and Aider were dropped; do not
+re-add adapters for them:
+- Claude Code, Antigravity, OpenCode, Mistral Vibe, Crush, Pi Coding Agent
+- Open WebUI, Charmbracelet Mods, or any tool that takes a system prompt string
+- Any other tool that reads `AGENTS.md`, which is the de facto convention

 The canonical source is always `.context/AGENT.md` (root) and `.context/PROJECT.md` (per-project).
 Derived files are committed (see *How context propagates* below) so a `git pull` on any host yields full agent context with no setup.
@@ -249,15 +291,29 @@ Canonical sources of truth:
 - Project: `<repo>/.context/PROJECT.md` (per-repo)

 Derived files (committed, regenerated by `task context:sync`):
- `CLAUDE.md`, `AGENTS.md`, `.cursorrules`, `.aider.conventions.md`,
-  `.context/system-prompt.txt`
+- `CLAUDE.md`, `AGENTS.md`, `.context/system-prompt.txt`
+
+Repos still carrying `.cursorrules`, `.aider.conventions.md` or
+`.aider.conf.yml` are stale: nothing regenerates them. Delete them the next time
+you touch the repo, along with their entries in that repo's `task check`
+assertion.

 Workflow:
 1. Edit a canonical file. Run `task context:sync`. Commit canonical and
   derived together. Push.
-2. On any other host, `git pull` brings both. Claude Code (tree-walking)
-   uses `CLAUDE.md`; Crush / Pi / Antigravity (cwd-only) use `AGENTS.md`;
-   Cursor uses `.cursorrules`; Aider uses `.aider.conventions.md`.
+2. On any other host, `git pull` brings both. Which file a harness reads does
+   NOT split cleanly by lookup strategy, so do not reason from the strategy:
+   - `CLAUDE.md` — Claude Code (walks up the tree)
+   - `AGENTS.md` — Crush, Pi, Antigravity (cwd-only), OpenCode (project root,
+     global at `~/.config/opencode/AGENTS.md`), and Mistral Vibe, which **walks
+     up** from cwd and takes the first `AGENTS.md` it finds
+     (user-level `~/.vibe/AGENTS.md`)
+   - `.context/system-prompt.txt` — Open WebUI, Mods, raw API calls; not read by
+     any harness, it is for pasting
+
+   Vibe tree-walks for `AGENTS.md`, so "tree-walkers use CLAUDE.md" is false.
+   Each repo's `AGENTS.md` embeds the canonical and is the first one found from
+   inside that repo, so this works — but by construction, not by luck.
 3. `task check` runs `context:sync` then asserts `git status --porcelain`
   is empty over the derived files (catches both modified-tracked drift
   and missing-untracked adapters). A drift fails the check with a
@@ -268,7 +324,20 @@ unconditionally on every host, every harness.

 ## Engineering Skills

-Shared engineering skills live in the **`mathias/skills`** repo (`git.d-ma.be/mathias/skills`). Clone it to `~/dev/skills/` and run `SKILLS_CHECKOUT_DIR="$PWD" bash install.sh` there to wire every skill into your harnesses (Claude Code, Crush, Antigravity, Mistral Vibe) as native, on-demand skills. (Use `install.sh`, not `task install` — the latter is currently broken, skills#7.) Load at task start — not "on demand" but on schedule, before writing code. Browse `~/dev/skills/SKILLS_INDEX.md` for the full list.
+Shared engineering skills live in the **`mathias/skills`** repo (`git.d-ma.be/mathias/skills`). Load at task start — not "on demand" but on schedule, before writing code. Browse `SKILLS_INDEX.md` for the full list.
+
+**Claude Code** installs the repo as a plugin from the marketplace the same repo hosts. That path also carries the `secret-guard` and `pre-task-ritual` hooks and the brain/gitea/infra MCP servers, which symlinks cannot:
+
+```
+/plugin marketplace add https://git.d-ma.be/mathias/skills.git
+/plugin install skills@mathias
+```
+
+Skills are **namespaced** under Claude Code: load `tdd` as `skills:tdd`. The trigger table below uses bare names — prefix them.
+
+**Crush, Pi and OpenCode** have no plugin system and are wired by `install.sh` as one directory symlink per harness via `~/.agents/skills`. Do not clone the checkout into `~/dev`: that folder is iCloud-shared between the Macs and spans two path roots (`/Users/mathias` vs `/home/mathias` on koala), so a checkout there is a git working tree with two concurrent writers, and any absolute symlink stored in it is broken on one side by construction (skills#23). Leave `SKILLS_CHECKOUT_DIR` at its host-local default.
+
+Check a host with `task doctor` in the skills repo; it also runs automatically at session start via the plugin.

 **Skill trigger table — load before starting, not after getting stuck:**

@@ -1,389 +0,0 @@
-# Cursor rules — auto-generated
-# Do not edit. Run: task context:sync
-
-# Agent context — Mathias workspace
-
-<!-- Canonical root context for all AI coding agents.
-     Lives at: ~/dev/.context/AGENT.md
-     Applies to every project under ~/dev/ unless overridden.
-     
-     Run `task context:sync` from ~/dev/ to regenerate harness-specific files.
-     Project-level context in .context/PROJECT.md layers on top of this. -->
-
-## Who I am
-
-I'm Mathias, a digital product manager and technology consultant based in Sweden.
-I build software, research emerging tech, and deliver consulting engagements
-for clients under NDA. I work across AI/ML, financial automation, web applications,
-and climate/sustainability tech.
-
-## How I work with agents
-
- I think like a product manager — I care about *why* before *how*
- I want agents to be opinionated and push back, not just execute blindly
- I prefer concise responses; skip ceremony and get to the point
- When I say "build this", I mean production-quality with tests, not a demo
- Ask me before making irreversible changes or adding heavy dependencies
- I work with confidential client data — never send it to cloud APIs unless I explicitly say it's OK
-
-## Behavior rules
-
-These rules apply to every task across every project, regardless of harness.
-
-0. **Pre-task ritual — before ANY implementation (non-negotiable).** Run this before writing a single line:
-   - **Query the brain** (`brain_query`) for the domain + symptom. If the result changes your approach, surface it before acting. 5 seconds beats 5 hours.
-   - **Load the relevant skill** — see trigger table in *Engineering Skills* below.
-   - **Write the failing test first.** Name the test before the function. If the target is untestable (e.g. `main()` wiring), extract the logic into a testable function first. No implementation without a red test.
-   - **State the observable success criterion** — what specific behavior, output, or passing test proves this is done?
-
-   **TDD is non-negotiable.** "Tests pass" is not proof of correctness — only proof the tests ran. Write tests that would catch the bug before writing code that fixes it.
-
-1. **No assumptions.** Don't hide confusion — surface it. Surface tradeoffs explicitly.
-   Think before coding; if the problem is unclear, ask or state assumptions before acting.
-2. **Minimum viable code.** Solve with the smallest change that works. Nothing
-   speculative, no "while we're here" cleanups, no premature abstractions. Simplicity first.
-3. **Surgical changes.** Touch only what the task requires. Leave unrelated code,
-   files, and formatting alone. Diffs should be small and reviewable.
-4. **Goal-driven execution.** Define clear success criteria up front for every task.
-   Loop — implement, verify, refine — until those criteria are met. Don't claim
-   completion without evidence (tests pass, command output, observed behavior).
-5. **Trunk-Based Development — commit directly to main.** Every commit is one
-   logical change (one tool, one fix, one test) with passing tests. Main is always
-   deployable. Never create long-lived feature branches.
-
-   **Exception — parallel agents on same repo:** If another agent is known to be
-   actively working on the same repo simultaneously, create a short-lived branch
-   (`agent/<description>`), finish the task, and merge to main within the same
-   session. Do not leave agent branches open between sessions.
-
-   **Exception — external contributor or client four-eyes requirement:** Use
-   PR flow only when a human reviewer outside the project is required. Document
-   the reason in PROJECT.md.
-
-6. **Close the loop — every substantive task ends with the same ritual.** Shipping
-   the code is not the end of the task; capturing it is. Run this unprompted:
-   - **Tag + bump SemVer** on the change (annotated tag; minor for a feature or
-     new/changed ADR, patch for a fix; docs in the same commit). Check the repo's
-     actual last tag — stated versions in docs drift stale.
-   - **Push** main and the tag (CI is the gate).
-   - **Persist generalizable learnings to the brain** (`brain_write`, wing/hall) —
-     the reusable patterns and the footguns that would bite anyone again, never
-     project status. See *Knowledge base — when to write* below.
-   - **File discovered-but-deferred work as tracker issues** on the project's own
-     repo — token-budget gaps, recorded ADR limitations, v2 follow-ups. Don't let
-     "out of scope, recorded" rot in a commit message; make it a ticket with a
-     source pointer.
-   - Surface the brain entries and issue numbers in the closing summary so the
-     trail is auditable.
-
-## Default stack
-
-| Layer | Default | Fallback | Last resort |
-|-------|---------|----------|-------------|
-| Language | Go | Python | TypeScript, Java, C |
-| UI | HTMX + Templ | Server-rendered HTML | React (only if SPA is justified) |
-| Build | Task (taskfile.dev) | Make | — |
-| Containers | Docker Compose (dev), k3s (prod) | — | — |
-| DB | PostgreSQL + sqlc | SQLite | — |
-| Search | pgvector (vector), BM25 | Qdrant (when >1M vectors or hybrid retrieval) | — |
-| Logging | slog (structured) | — | — |
-| Testing | Table-driven, testify | — | — |
-| Agents (Go) | google.golang.org/adk + pkg/litellm adapter | — | — |
-
-Exploratory: Rust, Zig — I'll tell you when I want these.
-
-## Code conventions
-
- **Go style**: golines, gofumpt, golangci-lint
- **Errors**: `fmt.Errorf("operation: %w", err)` — never naked, never log-and-return
- **Naming**: stdlib conventions, no stuttering
- **Architecture**: prefer stdlib over frameworks, constructor injection, env-var config parsed into typed structs
- **Git**: conventional commits (`feat:`, `fix:`, `chore:`), commit directly to main,
-  one logical change per commit, CI is the quality gate
- **Never**: long-lived feature branches, PRs for solo work, direct push without
-  passing `task check` locally first
- **Security**: no secrets in code, govulncheck before adding deps, SOPS for encrypted config
- **Dependencies**: prefer stdlib. testify, slog, templ, sqlc, google.golang.org/adk (agent projects only) are pre-approved; anything else needs justification in the commit message
-
-## Secret handling (every harness, every command)
-
-Tool output is persisted: terminal → `~/.claude/projects` transcripts →
-claudewatcher → brain/wiki → gitea history. A secret printed once is
-searchable forever, and clearing it means rotating the key. So:
-
-1. **Never print, echo, log, or transform a secret to inspect it.** No
-   `base64`/`xxd`/`cat` of a key, and never pipe a secret through a transform
-   to defeat `op run`'s output masking (it masks raw values; base64 hides them
-   from the mask — that exact trick leaked a key on 2026-06-11).
-2. **Secrets stay in the subprocess.** Reference them only as env vars consumed
-   *inside* `op run --env-file ~/.op-env -- <cmd>`. Never place a literal secret
-   in a command's argv (it lands in the tool call and the transcript).
-3. **Existence check without revealing the value:** `[ -n "$X" ] && echo set` —
-   never `${X:-...}` (returns the value when set) and never echo a substring of it.
-4. **Cross-host secrets:** run the secret-consuming command on the host that has
-   the secret; do not forward a raw key over ssh argv/stdout.
-5. If a secret does leak into output, say so immediately and flag it for rotation —
-   don't bury it.
-
-## Infrastructure
-
-Three machines on Tailscale:
-
-| Machine | Role | Key specs |
-|---------|------|-----------|
-| koala | GPU inference, heavy compute | RTX 5070, runs k3s + llama-swap + shared postgres18/pgvector |
-| iguana | Services, builds | M2 Ultra Mac |
-| flamingo | Daily driver, edge | Mac mini, ~/dev is here |
-
- **Model routing**: LiteLLM in front of llama-swap (local) + cloud APIs (when permitted)
- **Orchestration**: k3s cluster across all three machines
- **Networking**: Tailscale mesh
-
-## Project landscape
-
-All development repos live at `~/dev/` (softlink from `~/Documents/local-dev/`).
-
-Organized in thematic folders:
-
-| Folder | Focus | Count |
-|--------|-------|-------|
-| `GO/` | Go web frameworks, API integrations, learning projects | ~10 |
-| `AI/` | ML research, AI frameworks (FinRL, DSPy, crawl4ai) | ~6 |
-| `AGENTS/` | Autonomous agents, coding agents, MCP servers, infra | ~15 |
-| `QKX/` | Invoice processing, financial automation, payment systems | ~13 |
-| `XT/` | Climate data, sustainability (Klimatkollen, Garbo) | ~2 |
-
-See `~/dev/PROJECT_SUMMARY.md` for detailed descriptions of each project.
-
-### Key active projects
-
- **super-koala** (`AGENTS/`) — multi-component agent stack with LangGraph, DSPy, MCP
- **azure-tiger** (`QKX/`) — invoice extraction → ISO 20022 payment instructions
- **gocrwl** (`AGENTS/`) — Go web crawler with containerized deployment
- **koala-ai-stack** (`AGENTS/`) — local AI server infrastructure management
- **klimatkollen** (`XT/`) — Swedish municipal climate data platform
-
-## Knowledge base — actively use it
-
-A persistent brain (BM25 search + LLM-synthesised Q&A) survives across sessions,
-hosts, and harnesses. It holds 100+ hard-won entries: infra incident postmortems,
-Go pitfalls, framework gotchas, design principles, ADRs. **It is not optional
-reference material — query it actively, not just when explicitly told.**
-
-### When to query (treat as a reflex)
-
- **Before** starting a non-trivial task — search for prior art with the symptom
-  AND the system component ("how did we solve X in Y?"). 5 seconds beats 5 hours.
- **When debugging** — search for the error string, the stack frame, the affected
-  service. Past you may have already paid this tax.
- **Before adopting** a pattern, library, framework, or model name — check if it
-  was tried and rejected, or what the integration footguns are.
- **When making architectural decisions** — search for the domain + "ADR" or
-  "decision" to find prior reasoning before re-deriving it.
- **When a recommendation feels novel** — challenge yourself: "has this been
-  documented?" The brain often has it.
-
-### When to write
-
-After you discover something that **future-you would forget** and that **isn't
-recoverable from the code, git log, or PR description alone**:
-
- Bugs whose root cause is non-obvious and generalisable beyond this project.
- Framework / library / model-name quirks that bit you and would bite anyone.
- Design principles validated under fire (e.g. "every `_get` needs a `_list`").
- Postmortems for incidents: what broke, why, how diagnosed, what to do next time.
-
-DON'T write project status, sprint progress, PR summaries, or "what I did this
-session" — those rot fast and the originals are in git/gitea anyway. Brain
-entries that age well are about *why*, *how to avoid*, and *what to do when*.
-
-### How to access (per harness)
-
-| Harness | Query | Write |
-|---------|-------|-------|
-| **Claude Code, Claude Desktop** | `brain_query` (BM25), `brain_answer` (LLM-synth + sources) MCP tools | `brain_write` MCP tool |
-| **Crush, Pi, Antigravity, other MCP-capable** | same MCP server: `ingestion-brain` (via the `mcp__*_brain__*` namespace once authenticated) | same |
-| **Anything HTTP-only (curl, scripts)** | `POST https://brain-mcp.d-ma.be/query` with `{"query":"..."}` (auth via `BRAIN_MCP_TOKEN`) | `POST .../write` with `{"content":"...","filename":"..."}` |
-| **Browser / human inspection** | `https://git.d-ma.be/mathias/hyperguild` → `knowledge/` and `wiki/` markdown files |
-
- **Scoping**: defaults to `public` collection; client projects filter to `{client}` + `public`.
- **Routing**: brain_answer's LLM uses berget.ai as primary, iguana ollama as
-  fallback. Both are configurable in the `supervisor/ingestion-deployment.yaml`
-  on the koala k3s cluster; don't hardcode local-only model names into the
-  berget URL (see knowledge entry on namespace mismatches).
-
-### Quick reflex checks
-
-If you find yourself about to say any of these out loud, you owe yourself a brain query first:
-
- "I think the issue might be..."
- "Let me try X and see..."
- "I'll just write a script to..."
- "This is probably a new bug..."
- "Has anyone done this before?" — *yes, probably, go check.*
-
-## Client work rules
-
-When working on a project tagged with a client name:
-1. Never send code, data, or context to cloud APIs — use local models only
-2. Never reference other client projects or their data
-3. Keep all artifacts within the client's git org / directory
-4. Treat everything as confidential unless told otherwise
-
-## Harness-agnostic principles
-
-This context is designed to work with any AI coding tool:
- Claude Code, Cursor, Aider, Open WebUI, Charmbracelet Mods/Crush
- Pi Coding Agent, Mistral Vibe, Antigravity
- Any tool that accepts a system prompt or reads a markdown context file
-
-The canonical source is always `.context/AGENT.md` (root) and `.context/PROJECT.md` (per-project).
-Derived files are committed (see *How context propagates* below) so a `git pull` on any host yields full agent context with no setup.
-
-## How context propagates
-
-Canonical sources of truth:
- Universal: `~/dev/.context/AGENT.md` (this file)
- Project: `<repo>/.context/PROJECT.md` (per-repo)
-
-Derived files (committed, regenerated by `task context:sync`):
- `CLAUDE.md`, `AGENTS.md`, `.cursorrules`, `.aider.conventions.md`,
-  `.context/system-prompt.txt`
-
-Workflow:
-1. Edit a canonical file. Run `task context:sync`. Commit canonical and
-   derived together. Push.
-2. On any other host, `git pull` brings both. Claude Code (tree-walking)
-   uses `CLAUDE.md`; Crush / Pi / Antigravity (cwd-only) use `AGENTS.md`;
-   Cursor uses `.cursorrules`; Aider uses `.aider.conventions.md`.
-3. `task check` runs `context:sync` then asserts `git status --porcelain`
-   is empty over the derived files (catches both modified-tracked drift
-   and missing-untracked adapters). A drift fails the check with a
-   message telling you to stage the regenerated files.
-
-Behavior rules in this file and per-project rules in `PROJECT.md` apply
-unconditionally on every host, every harness.
-
-## Engineering Skills
-
-Shared engineering skills live in the **`mathias/skills`** repo (`git.d-ma.be/mathias/skills`). Clone it to `~/dev/skills/` and run `SKILLS_CHECKOUT_DIR="$PWD" bash install.sh` there to wire every skill into your harnesses (Claude Code, Crush, Antigravity, Mistral Vibe) as native, on-demand skills. (Use `install.sh`, not `task install` — the latter is currently broken, skills#7.) Load at task start — not "on demand" but on schedule, before writing code. Browse `~/dev/skills/SKILLS_INDEX.md` for the full list.
-
-**Skill trigger table — load before starting, not after getting stuck:**
-
-| Task type | Load |
-|-----------|------|
-| Any feature or bug fix | `tdd` |
-| Refactor or design | `clean-code` or `solid` |
-| Debug | `problem-analysis` |
-| Review code or PRs | `code-review` |
-| Frame a problem before coding | `problem-analysis` |
-
---
-
-# Project context
-
-<!-- Canonical project context. Edit this, run `task context:sync`.
-     Root agent context from ~/dev/.context/AGENT.md is automatically
-     prepended for harnesses that don't walk the directory tree. -->
-
-## Identity
-
- **Name**: gitea-mcp
- **Owner**: Mathias
- **Client**: personal
- **Repo**: https://git.d-ma.be/mathias/gitea-mcp
- **Status**: active
-
-## Stack
-
- **Primary language**: Go
- **UI layer**: HTMX + Templ (when applicable)
- **Fallback languages**: Python, TypeScript (justify in PR if used)
- **Build**: Task (taskfile.dev), not Make
- **Containers**: Docker (compose for dev, k3s for deploy)
- **Target infra**: koala (GPU workloads), iguana (services), flamingo (edge)
-
-## Conventions
-
-### Code style
- Go: follow `golines`, `gofumpt`, `golangci-lint` with project config
- Tests: table-driven, in `_test.go` next to source, `testify` for assertions
- Errors: wrap with `fmt.Errorf("operation: %w", err)`, no naked returns
- Naming: stdlib conventions, no stuttering (`http.Client` not `http.HTTPClient`)
-
-### Architecture preferences
- Prefer standard library over frameworks (net/http over gin/echo)
- Dependency injection via constructor functions, not containers
- Configuration via environment variables, parsed at startup into a typed struct
- Structured logging via `slog`
-
-### Git
- Conventional commits: `feat:`, `fix:`, `chore:`, `docs:`, `refactor:`
- **Trunk-Based Development:** commit directly to main. One logical change per commit.
- Run `task check` locally before every push. CI is the quality gate, not branch protection.
- No feature branches, no PRs for solo/agent work.
- Exception: if a parallel agent session is active on this repo, use a short-lived
-  `agent/<description>` branch and merge within the same session.
-
-### Security
- No secrets in code, ever — use env vars or SOPS-encrypted files
- Client data never leaves local network unless explicitly cleared
- Dependencies: audit with `govulncheck` before adding
-
-## Knowledge base access
-
-This project can query the shared knowledge base via MCP or HTTP:
-
- **MCP endpoint**: `mcp://localhost:3100/knowledge`
- **HTTP fallback**: `http://localhost:3100/api/v1/search`
- **Scoping**: queries are filtered to collection `personal` + `public`
-
-## Behavior rules
-
-These rules apply to every task in this project, regardless of harness.
-
-1. **No assumptions.** Don't hide confusion — surface it. Surface tradeoffs explicitly.
-   Think before coding; if the problem is unclear, ask or state assumptions before acting.
-2. **Minimum viable code.** Solve with the smallest change that works. Nothing
-   speculative, no "while we're here" cleanups, no premature abstractions. Simplicity first.
-3. **Surgical changes.** Touch only what the task requires. Leave unrelated code,
-   files, and formatting alone. Diffs should be small and reviewable.
-4. **Goal-driven execution.** Define clear success criteria up front for every task.
-   Loop — implement, verify, refine — until those criteria are met. Don't claim
-   completion without evidence (tests pass, command output, observed behavior).
-
-## Agent instructions
-
-When acting as a coding agent on this project:
-
-1. Read this file and all `SKILL.md` files in `.skills/` before starting work
-2. Run `task check` before committing (lint + test + vet)
-3. If unsure about a convention, check `DECISIONS.md` or ask
-4. Never modify files outside the project root without explicit permission
-5. When adding a dependency, explain why in the commit message
-6. Commit directly to main. Run `task check` before every push. Never create
-   feature branches unless a parallel agent is simultaneously active on this repo.
-7. For client projects: never send code or context to cloud APIs — use local models via LiteLLM
-
-## Current state — v0.2.5 (2026-05-17)
-
-All v0.2 work is complete and deployed. No active sprint.
-
-### What shipped
-
-| Tag | PR | Tools / fixes |
-|-----|----|---------------|
-| v0.2.2 | #21 | repo_create, repo_update, repo_mirror_push |
-| v0.2.3 | #22 | repo_tree, repo_topics_update, file_read dir fix |
-| v0.2.4 | #23 | issue_get, release_create, repo_delete |
-| v0.2.5 | #26 | repo_update archived+template, create_project_from_template template_name, pr_files_diff loop fix |
-
-Current main: `e31fd3f`. CI green. Deployed via Flux.
-
-### Next up
-
-1. **`hyperguild new-project` v1** — primary next target.
-   See brain node `adr-new-project-gitea-first-github-mirror` for full flow spec.
-
-2. **Issue #19** — end-to-end mirror flow verification.
-   `repo_mirror_push` is implemented but the full flow (create repo → add push mirror → verify sync to GitHub) has not been tested manually. Do this before relying on it in production.
@@ -0,0 +1,121 @@
+#!/usr/bin/env bash
+# Build an OCI image with buildctl and push it to the local registry.
+#
+# CANONICAL COPY: mathias/infra ci-templates/build-image.sh (infra#309).
+# Every repo carries a byte-identical copy at .gitea/build-image.sh, written by
+# `task ci:sync` and held in place by the drift gate. DO NOT EDIT THE COPIES --
+# edit here, re-sync, and the gate will tell you which repos still disagree.
+#
+# Why this exists: this step was hand-copied into a dozen repos, so migrating it
+# off rootless buildah (infra#241) meant eleven near-identical commits, eleven
+# verifications, and three separate rediscoveries of the same fix in the two
+# repo templates and the gitea-ci skill. The differences between the copies were
+# incidental -- a label here, a build-arg there -- never intentional. Those are
+# inputs now, not forks of the implementation.
+#
+# Why buildctl and not buildah: the gitea-runner pod runs with
+# `capabilities: drop [ALL]` and `allowPrivilegeEscalation: false`, which
+# unconditionally blocks the setuid-root exec that rootless buildah's uid/gid
+# remapping needs, and the pod has no OCI runtime at all. buildctl is a thin
+# client -- it hands the build to the host's buildkitd over a socket, so the
+# privileged layer work happens on the host and this pod's securityContext is
+# not in the picture. buildah is absent from the runner image as of baked-v14
+# and its absence is asserted at pod start (infra#308).
+#
+# Inputs, all via environment:
+#
+#   IMAGE            (required) image name, e.g. cobalt-dingo
+#   REGISTRY         default localhost:5000
+#   SHA_TAG          default `git rev-parse --short HEAD`
+#   VERSION_TAG      optional extra tag, e.g. v1.2.3 on a tag build
+#   PUSH_LATEST      default 1; set 0 to skip the :latest tag
+#   CONTEXT          default . -- build context directory
+#   DOCKERFILE_DIR   default $CONTEXT -- directory holding the Dockerfile
+#   DOCKERFILE_NAME  optional, when the file is not named `Dockerfile`
+#   TARGET           optional, multi-stage build target (--opt target=)
+#   BUILD_ARGS       optional, space-separated K=V pairs
+#   BUILD_SECRETS    optional, space-separated buildkit secret specs, each
+#                    `id=NAME,src=PATH`. Pass the SPEC only -- the caller
+#                    writes the secret to PATH itself. A secret VALUE must
+#                    never be passed here: it would land in this script's
+#                    argv and therefore in the CI log.
+#   EXTRA_LABELS     optional, space-separated K=V pairs
+#   GIT_SHA          optional, stamped as org.opencontainers.image.revision
+#   IMAGE_SOURCE     optional, stamped as org.opencontainers.image.source
+#
+# Note on shell style below: every conditional is a full `if` block, never
+# `[ -n "$X" ] && cmd`. Under `set -e` the && form exits the script whenever the
+# test is false, which is a real bug this estate has already shipped once.
+
+set -euo pipefail
+
+if [ -z "${IMAGE:-}" ]; then
+  echo "build-image: IMAGE is required" >&2
+  exit 2
+fi
+
+REGISTRY="${REGISTRY:-localhost:5000}"
+BUILDKIT_HOST="${BUILDKIT_HOST:-unix:///run/buildkit/buildkitd.sock}"
+CONTEXT="${CONTEXT:-.}"
+DOCKERFILE_DIR="${DOCKERFILE_DIR:-$CONTEXT}"
+PUSH_LATEST="${PUSH_LATEST:-1}"
+VERSION_TAG="${VERSION_TAG:-}"
+
+if [ -z "${SHA_TAG:-}" ]; then
+  SHA_TAG="$(git rev-parse --short HEAD)"
+fi
+
+REF="${REGISTRY}/${IMAGE}:${SHA_TAG}"
+
+TAR_FILE="$(mktemp)"
+trap 'rm -f "$TAR_FILE"' EXIT
+
+args=(
+  --frontend dockerfile.v0
+  --local "context=${CONTEXT}"
+  --local "dockerfile=${DOCKERFILE_DIR}"
+)
+
+if [ -n "${DOCKERFILE_NAME:-}" ]; then
+  args+=(--opt "filename=${DOCKERFILE_NAME}")
+fi
+if [ -n "${TARGET:-}" ]; then
+  args+=(--opt "target=${TARGET}")
+fi
+for spec in ${BUILD_SECRETS:-}; do
+  # Opaque id=NAME,src=PATH -- buildkit reads the file, nothing is echoed.
+  args+=(--secret "${spec}")
+done
+if [ -n "${GIT_SHA:-}" ]; then
+  args+=(--opt "label:org.opencontainers.image.revision=${GIT_SHA}")
+fi
+if [ -n "${IMAGE_SOURCE:-}" ]; then
+  args+=(--opt "label:org.opencontainers.image.source=${IMAGE_SOURCE}")
+fi
+for kv in ${BUILD_ARGS:-}; do
+  args+=(--opt "build-arg:${kv}")
+done
+for kv in ${EXTRA_LABELS:-}; do
+  args+=(--opt "label:${kv}")
+done
+
+echo "build-image: building ${REF}"
+buildctl --addr "$BUILDKIT_HOST" build "${args[@]}" --output "type=oci,dest=${TAR_FILE}"
+
+# buildctl has no registry-push equivalent to `buildah push`, so skopeo copies
+# the OCI archive up -- once per tag, from the same archive, so every tag points
+# at bit-identical content.
+push_tag() {
+  echo "build-image: pushing ${1}"
+  skopeo copy --dest-tls-verify=false "oci-archive:${TAR_FILE}" "docker://${1}"
+}
+
+push_tag "$REF"
+if [ "$PUSH_LATEST" = "1" ]; then
+  push_tag "${REGISTRY}/${IMAGE}:latest"
+fi
+if [ -n "$VERSION_TAG" ]; then
+  push_tag "${REGISTRY}/${IMAGE}:${VERSION_TAG}"
+fi
+
+echo "✓ build-image: ${REF} pushed"
@@ -58,41 +58,42 @@ jobs:
          REF="${{ github.ref }}"
          if [[ "$REF" == refs/tags/v* ]]; then
            echo "version-tag=${REF#refs/tags/}" >> "$GITHUB_OUTPUT"
+            echo "version=${REF#refs/tags/}" >> "$GITHUB_OUTPUT"
+          else
+            # The ldflags version stamp: git tag on a v* build, else short
+            # sha. Computed in shell rather than as a fallback expression in
+            # YAML, so it cannot depend on how the runner treats an empty step
+            # output -- a wrong VERSION would be baked into the binary silently.
+            #
+            # NOTE: no expression syntax in this comment, deliberately. act
+            # interpolates a run: block INCLUDING its comments, so prose
+            # containing that syntax is evaluated as an expression -- it failed
+            # with "Unavailable context: a", the step wrote no outputs, and the
+            # deploy job committed an empty image tag that broke Flux
+            # estate-wide (infra 51ceb7c).
+            echo "version=${SHA}" >> "$GITHUB_OUTPUT"
          fi

      - name: Build and push to local registry
-        run: |
-          REGISTRY="localhost:5000"
-          REF="${REGISTRY}/${{ env.IMAGE }}:${{ steps.meta.outputs.sha-tag }}"
-          # Stamp the real build version: the git tag on a v* build, else the short sha.
-          VERSION="${{ steps.meta.outputs.version-tag }}"
-          [ -z "$VERSION" ] && VERSION="${{ steps.meta.outputs.sha-tag }}"
-          buildah build \
-            --build-arg VERSION="${VERSION}" \
-            --label "org.opencontainers.image.revision=${{ github.sha }}" \
-            --label "org.opencontainers.image.source=${{ github.repositoryUrl }}" \
-            -t ${REF} \
-            -t ${REGISTRY}/${{ env.IMAGE }}:latest \
-            .
-          buildah push --tls-verify=false ${REF}
-          buildah push --tls-verify=false ${REGISTRY}/${{ env.IMAGE }}:latest
-          [[ -n "${{ steps.meta.outputs.version-tag }}" ]] && \
-            buildah push --tls-verify=false \
-              ${REF} \
-              ${REGISTRY}/${{ env.IMAGE }}:${{ steps.meta.outputs.version-tag }} || true
-          echo "✓ Image pushed to ${REF}"
+      # infra#309: implementation lives in .gitea/build-image.sh, a
+      # byte-identical copy of mathias/infra's ci-templates/build-image.sh.
+      # Do not edit it here -- edit the canonical file, run `task ci:sync`,
+      # and infra's drift gate names any repo still carrying an old copy.
+        env:
+          SHA_TAG: ${{ steps.meta.outputs.sha-tag }}
+          VERSION_TAG: ${{ steps.meta.outputs.version-tag }}
+          GIT_SHA: ${{ github.sha }}
+          IMAGE_SOURCE: ${{ github.repositoryUrl }}
+          BUILD_ARGS: VERSION=${{ steps.meta.outputs.version }}
+        run: bash .gitea/build-image.sh

-      - name: Smoke test
-        run: |
-          REGISTRY="localhost:5000"
-          REF="${REGISTRY}/${{ env.IMAGE }}:${{ steps.meta.outputs.sha-tag }}"
-          CNAME="smoke-${{ steps.meta.outputs.sha-tag }}"
-          sudo k3s ctr images pull --plain-http ${REF}
-          OUTPUT=$(timeout 5 sudo k3s ctr run --rm ${REF} ${CNAME} /gitea-mcp 2>&1 || true)
-          sudo k3s ctr containers delete ${CNAME} 2>/dev/null || true
-          echo "$OUTPUT" | grep -q "gitea-mcp" \
-            && echo "✓ Smoke test passed" \
-            || echo "⚠ Smoke test inconclusive: $OUTPUT"
+      # No smoke test here. The old one ran `sudo k3s ctr run` against the host's
+      # containerd: written for the Arch host runner, impossible in the
+      # containerized runner (infra#233) which has no sudo, no ctr and no
+      # containerd socket -- it failed with `sudo: command not found` the moment
+      # infra#241's buildctl migration let the build reach it. The deploy job's
+      # `Verify rollout` + `Confirm pod running new image` cover the same ground
+      # more strongly, which is why tapir dropped its smoke test too (infra#241).

  # ── 3. Deploy via infra repo + Flux ─────────────────────────────────────────
  deploy:
@@ -108,6 +109,13 @@ jobs:
          DEPLOY_KEY: ${{ secrets.INFRA_DEPLOY_KEY }}
        run: |
          set -euo pipefail
+          # infra#311: fail on an empty tag BEFORE patching anything. The
+          # grep verification further down CANNOT catch it -- with IMAGE_TAG
+          # empty its pattern matches the very corruption it exists to reject.
+          # The result is `image: <repo>:` in the manifest, which is invalid
+          # YAML, so Flux fails the ENTIRE apps Kustomization and every app
+          # stops reconciling, not just this one. That happened on 2026-08-25.
+          [ -n "${IMAGE_TAG:-}" ] || { echo "IMAGE_TAG is empty - refusing to patch"; exit 1; }

          mkdir -p ~/.ssh
          echo "$DEPLOY_KEY" > ~/.ssh/id_infra
@@ -48,14 +48,30 @@ These rules apply to every task across every project, regardless of harness.
   logical change (one tool, one fix, one test) with passing tests. Main is always
   deployable. Never create long-lived feature branches.

-   **Exception — parallel agents on same repo:** If another agent is known to be
-   actively working on the same repo simultaneously, create a short-lived branch
-   (`agent/<description>`), finish the task, and merge to main within the same
-   session. Do not leave agent branches open between sessions.
+   **The PR trigger is RISK, not concurrency.** `risk/low` work commits straight to
+   main with CI as the gate — that is most work and is unchanged. `risk/medium` and
+   above opens a PR, even solo, even with nobody waiting to review it. This mirrors
+   the tiered merge model in `assessor-loop`
+   `method/transfer-to-a-regulated-organisation.md` §4.1, so the estate practises
+   what it sells instead of merging everything on green and claiming otherwise.

-   **Exception — external contributor or client four-eyes requirement:** Use
-   PR flow only when a human reviewer outside the project is required. Document
-   the reason in PROJECT.md.
+   **Concurrency is NOT a reason to branch.** Measured 2026-08-29 across three real
+   collisions between parallel sessions: two were not branching problems at all — a
+   generated artifact written to the same path by both (branching turns that into a
+   merge conflict on a 33k-line JSON), and uncommitted work in a shared checkout
+   (a per-session worktree fixes it) — and the third was an ordinary non-fast-forward
+   push resolved by rebase in seconds. Use a worktree per session on shared repos and
+   timestamp generated artifacts. Do not reach for a branch.
+
+   Branch+PR flow carries documented costs on solo repos: merging auto-closes gated
+   issues, review comments live on a thread separate from the issue's, and a
+   merged-and-deleted branch cannot be amended (brain:
+   `gitea-pr-merge-autoclose-and-review-comment-footguns`). Those are the price of
+   the `risk/medium`+ lane — paid deliberately, not incurred by accident.
+
+   **Exception — external contributor or client four-eyes requirement:** Use PR flow
+   when a human reviewer outside the project is required. Document the reason in
+   PROJECT.md.

 6. **Close the loop — every substantive task ends with the same ritual.** Shipping
   the code is not the end of the task; capturing it is. Run this unprompted:
@@ -73,6 +89,26 @@ These rules apply to every task across every project, regardless of harness.
   - Surface the brain entries and issue numbers in the closing summary so the
     trail is auditable.

+   **Chat is not a durable store. Never rely on Mathias remembering anything.**
+   Every gap, caveat, risk, open question and next step must land in one of four
+   places before the turn ends: **the brain**, **a repo** (code, docs, ADR),
+   **a tracker issue**, or **a plan file**. A finding stated only in conversation is
+   lost the moment the session closes — and this estate has repeatedly proved that a
+   remembered convention is an unreliable control.
+
+   This applies hardest to what is tempting to leave as a closing remark: "worth
+   knowing X isn't enforced yet", "this is a convention, not a gate", "we should
+   revisit Y". Those are exactly the items that need a locator. If it is worth
+   saying in a summary, it is worth writing where it can be found again without you.
+
+## CAD sprint patterns
+
+Two CAD sprint archetypes (defs: brain `wiki/homelab/decisions/continuous-agentic-development-cad-concept-2026-06-16`):
+- **Inception Sprint** — birth a new repo. Oath-gated close; tag only once every clause holds. Brain: `.../inception-sprint-and-oath`.
+- **Sweep Sprint** — clear an existing backlog. Parallel read-only research per issue → sequential one-issue-per-commit execution. Brain: `.../sweep-sprint-and-oath`.
+
+Shared: decisions that are yours, not the agent's, surface as `DECISION NEEDED` (question/options/recommendation/blocks) — ask live if you're present, else comment and wait, never act past it. Live-breaking regressions get `git revert`, never forward-fixed; host-network work skips even read-only research (solo, synchronous, you present). Rule 6's closing ritual only fires once every Oath clause genuinely holds — blocked = descoped-and-tracked, never marked done.
+
 ## Default stack

 | Layer | Default | Fallback | Last resort |
@@ -83,7 +119,7 @@ These rules apply to every task across every project, regardless of harness.
 | Containers | Docker Compose (dev), k3s (prod) | — | — |
 | DB | PostgreSQL + sqlc | SQLite | — |
 | Search | pgvector (vector), BM25 | Qdrant (when >1M vectors or hybrid retrieval) | — |
-| Logging | slog (structured) | — | — |
+| Logging | slog (structured) | stdlib `logging` w/ structured `extra=` (or structlog) | — |
 | Testing | Table-driven, testify | — | — |
 | Agents (Go) | google.golang.org/adk + pkg/litellm adapter | — | — |

@@ -92,7 +128,12 @@ Exploratory: Rust, Zig — I'll tell you when I want these.
 ## Code conventions

 - **Go style**: golines, gofumpt, golangci-lint
- **Errors**: `fmt.Errorf("operation: %w", err)` — never naked, never log-and-return
+- **Python style** (fallback language): ruff (format+lint, one tool), mypy --strict (non-negotiable,
+  matches Go's static typing discipline), pytest + pytest-cov (table-driven via
+  `@pytest.mark.parametrize`), uv (venv+deps+lock, one tool), pydantic-settings (typed env-var config
+  — same principle as Go's typed structs), src-layout + `pyproject.toml` only (no `setup.py`)
+- **Errors**: `fmt.Errorf("operation: %w", err)` — never naked, never log-and-return.
+  Python: `raise X from e` (exception chaining, same principle) — never bare `except`, never silent `pass`
 - **Naming**: stdlib conventions, no stuttering
 - **Architecture**: prefer stdlib over frameworks, constructor injection, env-var config parsed into typed structs
 - **Git**: conventional commits (`feat:`, `fix:`, `chore:`), commit directly to main,
@@ -130,10 +171,10 @@ Three machines on Tailscale:
 |---------|------|-----------|
 | koala | GPU inference, heavy compute | RTX 5070, runs k3s + llama-swap + shared postgres18/pgvector |
 | iguana | Services, builds | M2 Ultra Mac |
-| flamingo | Daily driver, edge | Mac mini, ~/dev is here |
+| flamingo | Daily driver, edge | Apple Silicon Mac, ~/dev is here |

 - **Model routing**: LiteLLM in front of llama-swap (local) + cloud APIs (when permitted)
- **Orchestration**: k3s cluster across all three machines
+- **Orchestration**: k3s on koala only — single control-plane node, no workers (verified 2026-08-30). iguana and flamingo are on the tailnet but are not cluster members
 - **Networking**: Tailscale mesh

 ## Project landscape
@@ -229,10 +270,11 @@ When working on a project tagged with a client name:

 ## Harness-agnostic principles

-This context is designed to work with any AI coding tool:
- Claude Code, Cursor, Aider, Open WebUI, Charmbracelet Mods/Crush
- Pi Coding Agent, Mistral Vibe, Antigravity
- Any tool that accepts a system prompt or reads a markdown context file
+Supported harnesses (2026-09-01) — Cursor and Aider were dropped; do not
+re-add adapters for them:
+- Claude Code, Antigravity, OpenCode, Mistral Vibe, Crush, Pi Coding Agent
+- Open WebUI, Charmbracelet Mods, or any tool that takes a system prompt string
+- Any other tool that reads `AGENTS.md`, which is the de facto convention

 The canonical source is always `.context/AGENT.md` (root) and `.context/PROJECT.md` (per-project).
 Derived files are committed (see *How context propagates* below) so a `git pull` on any host yields full agent context with no setup.
@@ -244,15 +286,29 @@ Canonical sources of truth:
 - Project: `<repo>/.context/PROJECT.md` (per-repo)

 Derived files (committed, regenerated by `task context:sync`):
- `CLAUDE.md`, `AGENTS.md`, `.cursorrules`, `.aider.conventions.md`,
-  `.context/system-prompt.txt`
+- `CLAUDE.md`, `AGENTS.md`, `.context/system-prompt.txt`
+
+Repos still carrying `.cursorrules`, `.aider.conventions.md` or
+`.aider.conf.yml` are stale: nothing regenerates them. Delete them the next time
+you touch the repo, along with their entries in that repo's `task check`
+assertion.

 Workflow:
 1. Edit a canonical file. Run `task context:sync`. Commit canonical and
   derived together. Push.
-2. On any other host, `git pull` brings both. Claude Code (tree-walking)
-   uses `CLAUDE.md`; Crush / Pi / Antigravity (cwd-only) use `AGENTS.md`;
-   Cursor uses `.cursorrules`; Aider uses `.aider.conventions.md`.
+2. On any other host, `git pull` brings both. Which file a harness reads does
+   NOT split cleanly by lookup strategy, so do not reason from the strategy:
+   - `CLAUDE.md` — Claude Code (walks up the tree)
+   - `AGENTS.md` — Crush, Pi, Antigravity (cwd-only), OpenCode (project root,
+     global at `~/.config/opencode/AGENTS.md`), and Mistral Vibe, which **walks
+     up** from cwd and takes the first `AGENTS.md` it finds
+     (user-level `~/.vibe/AGENTS.md`)
+   - `.context/system-prompt.txt` — Open WebUI, Mods, raw API calls; not read by
+     any harness, it is for pasting
+
+   Vibe tree-walks for `AGENTS.md`, so "tree-walkers use CLAUDE.md" is false.
+   Each repo's `AGENTS.md` embeds the canonical and is the first one found from
+   inside that repo, so this works — but by construction, not by luck.
 3. `task check` runs `context:sync` then asserts `git status --porcelain`
   is empty over the derived files (catches both modified-tracked drift
   and missing-untracked adapters). A drift fails the check with a
@@ -263,7 +319,20 @@ unconditionally on every host, every harness.

 ## Engineering Skills

-Shared engineering skills live in the **`mathias/skills`** repo (`git.d-ma.be/mathias/skills`). Clone it to `~/dev/skills/` and run `SKILLS_CHECKOUT_DIR="$PWD" bash install.sh` there to wire every skill into your harnesses (Claude Code, Crush, Antigravity, Mistral Vibe) as native, on-demand skills. (Use `install.sh`, not `task install` — the latter is currently broken, skills#7.) Load at task start — not "on demand" but on schedule, before writing code. Browse `~/dev/skills/SKILLS_INDEX.md` for the full list.
+Shared engineering skills live in the **`mathias/skills`** repo (`git.d-ma.be/mathias/skills`). Load at task start — not "on demand" but on schedule, before writing code. Browse `SKILLS_INDEX.md` for the full list.
+
+**Claude Code** installs the repo as a plugin from the marketplace the same repo hosts. That path also carries the `secret-guard` and `pre-task-ritual` hooks and the brain/gitea/infra MCP servers, which symlinks cannot:
+
+```
+/plugin marketplace add https://git.d-ma.be/mathias/skills.git
+/plugin install skills@mathias
+```
+
+Skills are **namespaced** under Claude Code: load `tdd` as `skills:tdd`. The trigger table below uses bare names — prefix them.
+
+**Crush, Pi and OpenCode** have no plugin system and are wired by `install.sh` as one directory symlink per harness via `~/.agents/skills`. Do not clone the checkout into `~/dev`: that folder is iCloud-shared between the Macs and spans two path roots (`/Users/mathias` vs `/home/mathias` on koala), so a checkout there is a git working tree with two concurrent writers, and any absolute symlink stored in it is broken on one side by construction (skills#23). Leave `SKILLS_CHECKOUT_DIR` at its host-local default.
+
+Check a host with `task doctor` in the skills repo; it also runs automatically at session start via the plugin.

 **Skill trigger table — load before starting, not after getting stuck:**

@@ -27,12 +27,12 @@ tasks:
            echo "✓ context sync: skipped in CI"
          else
            bash scripts/context-sync.sh
-            drift=$(git status --porcelain -- AGENTS.md CLAUDE.md .cursorrules .aider.conventions.md .context/system-prompt.txt 2>/dev/null)
+            drift=$(git status --porcelain -- AGENTS.md CLAUDE.md .context/system-prompt.txt 2>/dev/null)
            if [ -n "$drift" ]; then
              echo "ERROR: derived adapters drifted from canonical context." >&2
              echo "$drift" >&2
              echo "" >&2
-              echo "Run: git add AGENTS.md CLAUDE.md .cursorrules .aider.conventions.md .context/system-prompt.txt" >&2
+              echo "Run: git add AGENTS.md CLAUDE.md .context/system-prompt.txt" >&2
              echo "     git commit -m 'chore: re-sync context adapters'" >&2
              exit 1
            fi
@@ -58,5 +58,3 @@ tasks:
    cmds: [bash scripts/context-sync.sh claude]
  context:sync:agents:
    cmds: [bash scripts/context-sync.sh agents]
-  context:sync:cursor:
-    cmds: [bash scripts/context-sync.sh cursor]
@@ -0,0 +1,77 @@
+// Command gitea-admin-mcp is a LOCAL-ONLY stdio MCP server exposing exactly
+// one tool: mint_service_account_token.
+//
+// This exists because dispatch#67 fix #2 hit a hard wall: Gitea's own
+// token-creation and admin-user endpoints reject token auth outright — an
+// existing API token cannot create another token, or a new account, by
+// design (a leaked token being usable to escalate would defeat the point of
+// scoping it). The only way through is real basic auth as an actual admin,
+// which a sandboxed agent session cannot safely hold. This binary is meant
+// to be spawned by a LOCAL Claude Code session (or run by hand) on a machine
+// with real, interactive 1Password access — configured as a stdio MCP
+// server in .mcp.json, or invoked directly.
+//
+// Deliberately NOT part of cmd/gitea-mcp (the shared HTTP server): the
+// capability here (reset a target account's password, mint it a token) is
+// far too dangerous to ever expose over the network, even behind OIDC.
+// internal/localtools is a separate package specifically so it can never be
+// wired into that server's RegisterAll by accident.
+//
+// Required env (fetch admin credentials just-in-time, e.g. via
+// `op run --env-file ~/.op-env -- gitea-admin-mcp`):
+//
+//	GITEA_BASE_URL        e.g. https://git.d-ma.be
+//	GITEA_ADMIN_USERNAME  a real site-admin account's username
+//	GITEA_ADMIN_PASSWORD  that account's real password
+//	OP_VAULT              1Password vault to store minted tokens in
+//	                      (default: AgentSecrets, per the estate's agent-governed
+//	                      ops-secrets policy)
+package main
+
+import (
+	"context"
+	"fmt"
+	"log/slog"
+	"os"
+
+	"git.d-ma.be/mathias/gitea-mcp/internal/adminbootstrap"
+	"git.d-ma.be/mathias/gitea-mcp/internal/localtools"
+	"git.d-ma.be/mathias/gitea-mcp/internal/mcp"
+	"git.d-ma.be/mathias/gitea-mcp/internal/registry"
+)
+
+func envOr(key, def string) string {
+	if v := os.Getenv(key); v != "" {
+		return v
+	}
+	return def
+}
+
+func main() {
+	// Logs to stderr, never stdout — stdout is the MCP JSON-RPC channel, and
+	// a stray log line there would corrupt the protocol stream.
+	logger := slog.New(slog.NewJSONHandler(os.Stderr, nil))
+
+	baseURL := os.Getenv("GITEA_BASE_URL")
+	adminUser := os.Getenv("GITEA_ADMIN_USERNAME")
+	adminPass := os.Getenv("GITEA_ADMIN_PASSWORD")
+	if baseURL == "" || adminUser == "" || adminPass == "" {
+		fmt.Fprintln(os.Stderr, "GITEA_BASE_URL, GITEA_ADMIN_USERNAME, and GITEA_ADMIN_PASSWORD are all required")
+		os.Exit(1)
+	}
+
+	b := adminbootstrap.Bootstrapper{
+		BaseURL:       baseURL,
+		AdminUsername: adminUser,
+		AdminPassword: adminPass,
+		Secrets:       localtools.OpSecretStore{Vault: envOr("OP_VAULT", "AgentSecrets")},
+	}
+
+	reg := registry.New()
+	reg.Register(localtools.NewMintServiceAccountToken(b))
+
+	if err := mcp.ServeStdio(context.Background(), reg, os.Stdin, os.Stdout); err != nil {
+		logger.Error("serve stdio", "err", err)
+		os.Exit(1)
+	}
+}
@@ -0,0 +1,92 @@
+# gitea-admin-mcp
+
+A separate, local-only stdio MCP server. One tool: `mint_service_account_token`.
+
+## Why this exists
+
+dispatch#67 fix #2 needed a read-only CI credential for a dedicated service
+account, and hit a hard wall: Gitea's own `POST /users/{u}/tokens` and
+`POST /admin/users` both reject token auth outright —
+
+    {"message":"auth required"}
+    {"message":"token does not have at least one of required scope(s), required=[write:admin], ..."}
+
+— by design: an existing API token cannot mint another token, or create an
+account, no matter its scope. The only way through either endpoint is real
+basic auth as an actual site admin. A sandboxed agent session has no
+interactive password and shouldn't be handed one — that's broader exposure
+than the job needs. This binary is meant to run where real admin credentials
+already exist just-in-time: your own machine, via `op run`.
+
+## What it does
+
+`mint_service_account_token` does the whole bootstrap in one call:
+
+1. Creates the named account if it doesn't exist (admin basic auth), or
+   resets its password if it does. **This is why `confirm_service_account`
+   is required and checked first** — the tool unconditionally (re)sets
+   whatever password the target account has, which is correct for a
+   dedicated bot account and would be destructive pointed at a real person's
+   login by mistake.
+2. Mints a scoped token for that account, authenticating as the account
+   itself (Gitea requires this specifically — an admin cannot mint a token
+   on another account's behalf).
+3. Grants the account read-only collaborator access on any named
+   `owner/repo` targets (admin action, least privilege).
+4. Stores the token in 1Password via `op item create`, value passed on
+   stdin — never as a literal argv element, matching the estate's own
+   secret-handling rule. Returns only the `op://` reference, never the
+   token value itself, to the caller.
+
+## Setup
+
+Deliberately **not** wired into `cmd/gitea-mcp` (the shared HTTP server) —
+this capability is too dangerous to expose over the network, even behind
+OIDC. `internal/localtools` is a separate package specifically so it can
+never be wired into that server's `RegisterAll` by accident.
+
+Build:
+
+```
+go build -o gitea-admin-mcp ./cmd/gitea-admin-mcp
+```
+
+Configure as a local stdio MCP server (Claude Code `.mcp.json`, run where
+your real `op` session lives):
+
+```json
+{
+  "mcpServers": {
+    "gitea-admin": {
+      "command": "op",
+      "args": ["run", "--env-file", "~/.op-env", "--", "/path/to/gitea-admin-mcp"]
+    }
+  }
+}
+```
+
+Required env (see `~/.op-env`):
+
+| Var | Meaning |
+|---|---|
+| `GITEA_BASE_URL` | e.g. `https://git.d-ma.be` |
+| `GITEA_ADMIN_USERNAME` | a real site-admin account |
+| `GITEA_ADMIN_PASSWORD` | that account's real password |
+| `OP_VAULT` | 1Password vault for minted tokens (default `AgentSecrets`, per the estate's agent-governed ops-secrets policy) |
+
+## Example call
+
+```json
+{
+  "username": "dispatch-bot",
+  "email": "dispatch@d-ma.be",
+  "scopes": ["read:repository", "read:issue"],
+  "token_name": "dispatch-ci-readonly",
+  "secret_name": "gitea-dispatch-bot-ci-readonly",
+  "grant_read_repos": ["mathias/cobalt-dingo"],
+  "confirm_service_account": true
+}
+```
+
+Result carries `SecretRef` (e.g. `op://AgentSecrets/gitea-dispatch-bot-ci-readonly/credential`)
+and `GrantedRepos` — never the token or either password.
@@ -0,0 +1,284 @@
+// Package adminbootstrap mints a scoped Gitea access token for a dedicated
+// service/bot account, end to end, in one auditable call.
+//
+// dispatch#67 fix #2 needed a read-only CI credential and hit a hard wall:
+// Gitea's own token-creation and admin-user endpoints reject token auth
+// outright (POST /users/{u}/tokens and POST /admin/users both return
+// "auth required" / a required-scope error for an existing token, by
+// design — a leaked token being usable to mint further tokens would be a
+// privilege-escalation footgun). Neither endpoint has a token-scoped
+// workaround. The only way through is real basic auth as an actual admin.
+//
+// That is exactly what a sandboxed agent session cannot safely have — no
+// interactive password, and handing one to a session would be broader
+// exposure than the job needs. This package is meant to run LOCALLY, where
+// real admin credentials exist just-in-time (1Password, `op run`), as a
+// small stdio MCP tool (see cmd/gitea-admin-mcp) rather than ad hoc curl.
+package adminbootstrap
+
+import (
+	"bytes"
+	"context"
+	"crypto/rand"
+	"encoding/base64"
+	"encoding/json"
+	"fmt"
+	"io"
+	"net/http"
+	"strings"
+)
+
+// HTTPDoer is the one method this package needs from *http.Client — narrow
+// on purpose, so tests inject a fake transport instead of hitting a network.
+type HTTPDoer interface {
+	Do(req *http.Request) (*http.Response, error)
+}
+
+// SecretStore persists the minted token somewhere durable and returns a
+// reference safe to log or display — never the value itself. The real
+// implementation (cmd/gitea-admin-mcp) shells out to the local `op` CLI;
+// tests use an in-memory fake.
+type SecretStore interface {
+	Store(ctx context.Context, name, value string) (ref string, err error)
+}
+
+// RepoRef names a repo to grant the new account read-only collaborator
+// access to — least privilege: a service account should see only the repos
+// it actually operates against, not every private repo on the instance.
+type RepoRef struct {
+	Owner string
+	Name  string
+}
+
+// Request describes one bootstrap. ConfirmServiceAccount is required and
+// checked first: this package (re)sets whatever password Username already
+// has, which is correct and safe for a dedicated bot account and dangerous
+// for anything else.
+type Request struct {
+	Username              string
+	Email                 string
+	FullName              string
+	Scopes                []string
+	TokenName             string
+	SecretName            string
+	GrantReadRepos        []RepoRef
+	ConfirmServiceAccount bool
+}
+
+// Result reports what happened. SecretRef, never the token value, is what a
+// caller (or an MCP tool's text response) should ever surface.
+type Result struct {
+	Username     string
+	Created      bool // false if the account already existed (password was reset, not created)
+	SecretRef    string
+	GrantedRepos []string
+}
+
+// Bootstrapper holds the real admin credentials and where to reach Gitea —
+// constructed once per invocation, never persisted.
+type Bootstrapper struct {
+	BaseURL       string
+	AdminUsername string
+	AdminPassword string
+	HTTP          HTTPDoer // nil defaults to http.DefaultClient
+	Secrets       SecretStore
+}
+
+func (b Bootstrapper) httpClient() HTTPDoer {
+	if b.HTTP != nil {
+		return b.HTTP
+	}
+	return http.DefaultClient
+}
+
+// Run creates (or resets the password of) Username, mints it a scoped
+// token, grants it read-only access to GrantReadRepos, stores the token via
+// Secrets, and returns without ever exposing the token value or either
+// password to the caller.
+func (b Bootstrapper) Run(ctx context.Context, req Request) (Result, error) {
+	if !req.ConfirmServiceAccount {
+		return Result{}, fmt.Errorf("refusing: ConfirmServiceAccount must be true — this resets %q's password unconditionally, which is only safe for a dedicated service/bot account", req.Username)
+	}
+
+	created, err := b.userExists(ctx, req.Username)
+	if err != nil {
+		return Result{}, fmt.Errorf("check existing account: %w", err)
+	}
+	created = !created // userExists returns true if found; Result.Created means "we created it"
+
+	botPassword, err := randomPassword()
+	if err != nil {
+		return Result{}, fmt.Errorf("generate account password: %w", err)
+	}
+
+	if created {
+		if err := b.createUser(ctx, req, botPassword); err != nil {
+			return Result{}, fmt.Errorf("create account: %w", err)
+		}
+	} else {
+		if err := b.resetPassword(ctx, req.Username, botPassword); err != nil {
+			return Result{}, fmt.Errorf("reset existing account's password: %w", err)
+		}
+	}
+
+	token, err := b.mintToken(ctx, req.Username, botPassword, req.TokenName, req.Scopes)
+	if err != nil {
+		return Result{}, fmt.Errorf("mint token: %w", err)
+	}
+
+	var granted []string
+	for _, r := range req.GrantReadRepos {
+		if err := b.grantReadAccess(ctx, r, req.Username); err != nil {
+			return Result{}, fmt.Errorf("grant read access on %s/%s: %w", r.Owner, r.Name, err)
+		}
+		granted = append(granted, r.Owner+"/"+r.Name)
+	}
+
+	ref, err := b.Secrets.Store(ctx, req.SecretName, token)
+	if err != nil {
+		return Result{}, fmt.Errorf("store token: %w", err)
+	}
+
+	return Result{
+		Username:     req.Username,
+		Created:      created,
+		SecretRef:    ref,
+		GrantedRepos: granted,
+	}, nil
+}
+
+func randomPassword() (string, error) {
+	raw := make([]byte, 32)
+	if _, err := rand.Read(raw); err != nil {
+		return "", err
+	}
+	return base64.RawURLEncoding.EncodeToString(raw), nil
+}
+
+func (b Bootstrapper) userExists(ctx context.Context, username string) (bool, error) {
+	resp, err := b.doAdmin(ctx, http.MethodGet, "/api/v1/users/"+username, nil)
+	if err != nil {
+		return false, err
+	}
+	defer func() { _ = resp.Body.Close() }()
+	switch resp.StatusCode {
+	case http.StatusOK:
+		return true, nil
+	case http.StatusNotFound:
+		return false, nil
+	default:
+		return false, fmt.Errorf("unexpected status %d checking account existence", resp.StatusCode)
+	}
+}
+
+func (b Bootstrapper) createUser(ctx context.Context, req Request, password string) error {
+	body := map[string]any{
+		"username":             req.Username,
+		"email":                req.Email,
+		"password":             password,
+		"must_change_password": false,
+	}
+	if req.FullName != "" {
+		body["full_name"] = req.FullName
+	}
+	resp, err := b.doAdmin(ctx, http.MethodPost, "/api/v1/admin/users", body)
+	if err != nil {
+		return err
+	}
+	defer func() { _ = resp.Body.Close() }()
+	if resp.StatusCode/100 != 2 {
+		return statusErr(resp)
+	}
+	return nil
+}
+
+func (b Bootstrapper) resetPassword(ctx context.Context, username, password string) error {
+	resp, err := b.doAdmin(ctx, http.MethodPatch, "/api/v1/admin/users/"+username, map[string]any{
+		"password": password,
+	})
+	if err != nil {
+		return err
+	}
+	defer func() { _ = resp.Body.Close() }()
+	if resp.StatusCode/100 != 2 {
+		return statusErr(resp)
+	}
+	return nil
+}
+
+// mintToken authenticates as the SERVICE ACCOUNT itself (not the admin) —
+// Gitea's token-creation endpoint requires basic auth as the token's owner,
+// and an admin's own credentials cannot mint on another account's behalf.
+func (b Bootstrapper) mintToken(ctx context.Context, username, password, tokenName string, scopes []string) (string, error) {
+	raw, err := json.Marshal(map[string]any{"name": tokenName, "scopes": scopes})
+	if err != nil {
+		return "", err
+	}
+	req, err := http.NewRequestWithContext(ctx, http.MethodPost,
+		strings.TrimRight(b.BaseURL, "/")+"/api/v1/users/"+username+"/tokens", bytes.NewReader(raw))
+	if err != nil {
+		return "", err
+	}
+	req.SetBasicAuth(username, password)
+	req.Header.Set("Content-Type", "application/json")
+
+	resp, err := b.httpClient().Do(req)
+	if err != nil {
+		return "", err
+	}
+	defer func() { _ = resp.Body.Close() }()
+	if resp.StatusCode/100 != 2 {
+		return "", statusErr(resp)
+	}
+
+	var parsed struct {
+		SHA1 string `json:"sha1"`
+	}
+	if err := json.NewDecoder(resp.Body).Decode(&parsed); err != nil {
+		return "", fmt.Errorf("decode token response: %w", err)
+	}
+	if parsed.SHA1 == "" {
+		return "", fmt.Errorf("token response carried no sha1 value")
+	}
+	return parsed.SHA1, nil
+}
+
+// grantReadAccess is an admin action (adding a collaborator to a repo the
+// admin owns), so it uses admin auth, not the new account's own.
+func (b Bootstrapper) grantReadAccess(ctx context.Context, r RepoRef, username string) error {
+	path := fmt.Sprintf("/api/v1/repos/%s/%s/collaborators/%s", r.Owner, r.Name, username)
+	resp, err := b.doAdmin(ctx, http.MethodPut, path, map[string]any{"permission": "read"})
+	if err != nil {
+		return err
+	}
+	defer func() { _ = resp.Body.Close() }()
+	if resp.StatusCode/100 != 2 {
+		return statusErr(resp)
+	}
+	return nil
+}
+
+func (b Bootstrapper) doAdmin(ctx context.Context, method, path string, body map[string]any) (*http.Response, error) {
+	var reader io.Reader
+	if body != nil {
+		raw, err := json.Marshal(body)
+		if err != nil {
+			return nil, err
+		}
+		reader = bytes.NewReader(raw)
+	}
+	req, err := http.NewRequestWithContext(ctx, method, strings.TrimRight(b.BaseURL, "/")+path, reader)
+	if err != nil {
+		return nil, err
+	}
+	req.SetBasicAuth(b.AdminUsername, b.AdminPassword)
+	if body != nil {
+		req.Header.Set("Content-Type", "application/json")
+	}
+	return b.httpClient().Do(req)
+}
+
+func statusErr(resp *http.Response) error {
+	msg, _ := io.ReadAll(resp.Body)
+	return fmt.Errorf("gitea returned %d: %s", resp.StatusCode, strings.TrimSpace(string(msg)))
+}
@@ -0,0 +1,258 @@
+package adminbootstrap_test
+
+import (
+	"context"
+	"encoding/json"
+	"io"
+	"net/http"
+	"strings"
+	"testing"
+
+	"git.d-ma.be/mathias/gitea-mcp/internal/adminbootstrap"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// This exists to close dispatch#67 fix #2's blocker: minting a scoped Gitea
+// token needs basic auth (POST /users/{u}/tokens and POST /admin/users both
+// reject token auth outright, confirmed live against git.d-ma.be — an
+// existing token cannot mint another token, by design). A sandboxed agent
+// session has no interactive password, so this package is meant to run
+// locally, where a real admin password is available just-in-time (1Password,
+// `op run`), and to do the whole bootstrap in one auditable call rather than
+// leaving it as ad hoc curl.
+//
+// Design choice, made explicit rather than silently assumed: this tool
+// ALWAYS (re)sets the target account's password via admin basic auth before
+// minting, whether the account is new or already existed. That is safe and
+// expected for a dedicated service/bot account (its password is a mechanism
+// to reach the token endpoint, never a human login credential) and would be
+// dangerous pointed at a real person's account by mistake -- which is why
+// Request.ConfirmServiceAccount exists and Run refuses to proceed without it.
+
+type call struct {
+	method, path string
+	body         map[string]any
+	authUser     string // basic-auth username used for this call, "" if token/none
+	authPass     string
+}
+
+type fakeTransport struct {
+	t         *testing.T
+	responses map[string]fakeResponse // key: "METHOD path"
+	calls     []call
+}
+
+type fakeResponse struct {
+	status int
+	body   string
+}
+
+func (f *fakeTransport) RoundTrip(req *http.Request) (*http.Response, error) {
+	var body map[string]any
+	if req.Body != nil {
+		raw, _ := io.ReadAll(req.Body)
+		if len(raw) > 0 {
+			require.NoError(f.t, json.Unmarshal(raw, &body))
+		}
+	}
+	user, pass, _ := req.BasicAuth()
+	f.calls = append(f.calls, call{method: req.Method, path: req.URL.Path, body: body, authUser: user, authPass: pass})
+
+	key := req.Method + " " + req.URL.Path
+	r, ok := f.responses[key]
+	if !ok {
+		f.t.Fatalf("unscripted request: %s (have: %v)", key, keys(f.responses))
+	}
+	return &http.Response{
+		StatusCode: r.status,
+		Body:       io.NopCloser(strings.NewReader(r.body)),
+		Header:     make(http.Header),
+	}, nil
+}
+
+func keys(m map[string]fakeResponse) []string {
+	out := make([]string, 0, len(m))
+	for k := range m {
+		out = append(out, k)
+	}
+	return out
+}
+
+// fakeSecretStore records what would have been written to 1Password, without
+// ever touching it -- and, critically, asserts the token value itself never
+// leaks into anything but this one call.
+type fakeSecretStore struct {
+	stored map[string]string
+}
+
+func (f *fakeSecretStore) Store(_ context.Context, name, value string) (string, error) {
+	if f.stored == nil {
+		f.stored = map[string]string{}
+	}
+	f.stored[name] = value
+	return "op://AgentSecrets/" + name + "/token", nil
+}
+
+func newBootstrapper(t *testing.T, responses map[string]fakeResponse) (adminbootstrap.Bootstrapper, *fakeTransport, *fakeSecretStore) {
+	ft := &fakeTransport{t: t, responses: responses}
+	fs := &fakeSecretStore{}
+	return adminbootstrap.Bootstrapper{
+		BaseURL:       "http://fake.gitea",
+		AdminUsername: "mathias",
+		AdminPassword: "admin-pass-not-a-real-secret",
+		HTTP:          &http.Client{Transport: ft},
+		Secrets:       fs,
+	}, ft, fs
+}
+
+func TestRun_CreatesNewAccountMintsTokenAndStoresIt(t *testing.T) {
+	b, ft, fs := newBootstrapper(t, map[string]fakeResponse{
+		"GET /api/v1/users/dispatch-bot":         {status: 404, body: `{}`},
+		"POST /api/v1/admin/users":               {status: 201, body: `{"login":"dispatch-bot"}`},
+		"POST /api/v1/users/dispatch-bot/tokens": {status: 201, body: `{"sha1":"tok_abc123"}`},
+	})
+
+	result, err := b.Run(context.Background(), adminbootstrap.Request{
+		Username:              "dispatch-bot",
+		Email:                 "dispatch@d-ma.be",
+		Scopes:                []string{"read:repository", "read:issue"},
+		TokenName:             "dispatch-ci-readonly",
+		SecretName:            "gitea-dispatch-bot-ci-readonly",
+		ConfirmServiceAccount: true,
+	})
+
+	require.NoError(t, err)
+	assert.True(t, result.Created)
+	assert.Equal(t, "op://AgentSecrets/gitea-dispatch-bot-ci-readonly/token", result.SecretRef)
+	assert.Equal(t, "tok_abc123", fs.stored["gitea-dispatch-bot-ci-readonly"])
+
+	// The account-creation call must use the ADMIN's basic auth, never the
+	// fresh account's own (unknown at that point) password.
+	create := findCall(t, ft.calls, "POST", "/api/v1/admin/users")
+	assert.Equal(t, "mathias", create.authUser)
+	assert.Equal(t, "dispatch-bot", create.body["username"])
+	assert.Equal(t, "dispatch@d-ma.be", create.body["email"])
+	assert.Equal(t, false, create.body["must_change_password"])
+
+	// The token-mint call must use basic auth AS THE NEW ACCOUNT -- Gitea
+	// requires this specifically (an admin token cannot mint on someone
+	// else's behalf) -- with the password this package generated and set,
+	// never the admin's own password.
+	mint := findCall(t, ft.calls, "POST", "/api/v1/users/dispatch-bot/tokens")
+	assert.Equal(t, "dispatch-bot", mint.authUser)
+	assert.NotEqual(t, "admin-pass-not-a-real-secret", mint.authPass)
+	assert.NotEmpty(t, mint.authPass)
+	assert.Equal(t, "dispatch-ci-readonly", mint.body["name"])
+	assert.ElementsMatch(t, []any{"read:repository", "read:issue"}, mint.body["scopes"])
+}
+
+func TestRun_ExistingAccountGetsPasswordResetNotRecreated(t *testing.T) {
+	b, ft, _ := newBootstrapper(t, map[string]fakeResponse{
+		"GET /api/v1/users/dispatch-bot":         {status: 200, body: `{"login":"dispatch-bot"}`},
+		"PATCH /api/v1/admin/users/dispatch-bot": {status: 200, body: `{}`},
+		"POST /api/v1/users/dispatch-bot/tokens": {status: 201, body: `{"sha1":"tok_new"}`},
+	})
+
+	result, err := b.Run(context.Background(), adminbootstrap.Request{
+		Username:              "dispatch-bot",
+		Email:                 "dispatch@d-ma.be",
+		Scopes:                []string{"read:repository"},
+		TokenName:             "dispatch-ci-readonly",
+		SecretName:            "gitea-dispatch-bot-ci-readonly",
+		ConfirmServiceAccount: true,
+	})
+
+	require.NoError(t, err)
+	assert.False(t, result.Created, "must report the account as pre-existing, not newly created")
+
+	// Must PATCH (reset password), never POST (create) for an existing account.
+	assert.Empty(t, filterCalls(ft.calls, "POST", "/api/v1/admin/users"))
+	patch := findCall(t, ft.calls, "PATCH", "/api/v1/admin/users/dispatch-bot")
+	assert.Equal(t, "mathias", patch.authUser, "the reset must be done as the real admin")
+	assert.NotEmpty(t, patch.body["password"])
+}
+
+func TestRun_RefusesWithoutExplicitServiceAccountConfirmation(t *testing.T) {
+	// Guard against pointing this at a real person's account by accident --
+	// this tool resets whatever password the target account has.
+	b, ft, _ := newBootstrapper(t, map[string]fakeResponse{})
+
+	_, err := b.Run(context.Background(), adminbootstrap.Request{
+		Username:   "mathias",
+		Email:      "mthbqv@gmail.com",
+		Scopes:     []string{"read:repository"},
+		TokenName:  "whatever",
+		SecretName: "whatever",
+		// ConfirmServiceAccount left false/unset
+	})
+
+	require.Error(t, err)
+	assert.Contains(t, err.Error(), "ConfirmServiceAccount")
+	assert.Empty(t, ft.calls, "must not make any request at all without confirmation")
+}
+
+func TestRun_GrantsReadOnlyCollaboratorAccessViaAdminAuth(t *testing.T) {
+	b, ft, _ := newBootstrapper(t, map[string]fakeResponse{
+		"GET /api/v1/users/dispatch-bot":                                    {status: 404, body: `{}`},
+		"POST /api/v1/admin/users":                                          {status: 201, body: `{}`},
+		"POST /api/v1/users/dispatch-bot/tokens":                            {status: 201, body: `{"sha1":"tok"}`},
+		"PUT /api/v1/repos/mathias/cobalt-dingo/collaborators/dispatch-bot": {status: 204, body: ``},
+	})
+
+	result, err := b.Run(context.Background(), adminbootstrap.Request{
+		Username:              "dispatch-bot",
+		Email:                 "dispatch@d-ma.be",
+		Scopes:                []string{"read:repository"},
+		TokenName:             "dispatch-ci-readonly",
+		SecretName:            "gitea-dispatch-bot-ci-readonly",
+		GrantReadRepos:        []adminbootstrap.RepoRef{{Owner: "mathias", Name: "cobalt-dingo"}},
+		ConfirmServiceAccount: true,
+	})
+
+	require.NoError(t, err)
+	assert.Equal(t, []string{"mathias/cobalt-dingo"}, result.GrantedRepos)
+	grant := findCall(t, ft.calls, "PUT", "/api/v1/repos/mathias/cobalt-dingo/collaborators/dispatch-bot")
+	assert.Equal(t, "mathias", grant.authUser, "granting collaborator access is an admin action, not the new account's own")
+	assert.Equal(t, "read", grant.body["permission"])
+}
+
+func TestRun_FailsClosedOnUpstreamError(t *testing.T) {
+	b, _, fs := newBootstrapper(t, map[string]fakeResponse{
+		"GET /api/v1/users/dispatch-bot": {status: 404, body: `{}`},
+		"POST /api/v1/admin/users":       {status: 500, body: `{"message":"internal error"}`},
+	})
+
+	_, err := b.Run(context.Background(), adminbootstrap.Request{
+		Username:              "dispatch-bot",
+		Email:                 "dispatch@d-ma.be",
+		Scopes:                []string{"read:repository"},
+		TokenName:             "x",
+		SecretName:            "x",
+		ConfirmServiceAccount: true,
+	})
+
+	require.Error(t, err)
+	assert.Empty(t, fs.stored, "nothing must be written to the secret store if the flow didn't complete")
+}
+
+func findCall(t *testing.T, calls []call, method, path string) call {
+	t.Helper()
+	for _, c := range calls {
+		if c.method == method && c.path == path {
+			return c
+		}
+	}
+	t.Fatalf("no call found: %s %s (have: %+v)", method, path, calls)
+	return call{}
+}
+
+func filterCalls(calls []call, method, path string) []call {
+	var out []call
+	for _, c := range calls {
+		if c.method == method && c.path == path {
+			out = append(out, c)
+		}
+	}
+	return out
+}
@@ -1,6 +1,11 @@
 package allowlist

-import "fmt"
+import (
+	"context"
+	"fmt"
+
+	"git.d-ma.be/mathias/gitea-mcp/internal/gitea"
+)

 type Allowlist struct {
 	owners map[string]struct{}
@@ -14,10 +19,17 @@ func New(owners []string) *Allowlist {
 	return &Allowlist{owners: m}
 }

-func (a *Allowlist) Check(owner string) error {
+// Check gates owner access to the static list — except for a caller
+// authenticated with their own Gitea PAT (pass-through, gitea-mcp#59), whose
+// access Gitea's own permission model already gates more precisely than a
+// coarse owner name list ever could.
+func (a *Allowlist) Check(ctx context.Context, owner string) error {
 	if owner == "" {
 		return fmt.Errorf("owner required")
 	}
+	if _, ok := gitea.TokenFromContext(ctx); ok {
+		return nil
+	}
 	if _, ok := a.owners[owner]; !ok {
 		return fmt.Errorf("owner %q not in allowlist", owner)
 	}
@@ -1,16 +1,36 @@
 package allowlist_test

 import (
+	"context"
 	"testing"

 	"git.d-ma.be/mathias/gitea-mcp/internal/allowlist"
+	"git.d-ma.be/mathias/gitea-mcp/internal/gitea"
 	"github.com/stretchr/testify/assert"
 )

 func TestAllowlistCheck(t *testing.T) {
 	a := allowlist.New([]string{"mathias", "acme"})
-	assert.NoError(t, a.Check("mathias"))
-	assert.NoError(t, a.Check("acme"))
-	assert.Error(t, a.Check("evil"))
-	assert.Error(t, a.Check(""))
+	ctx := context.Background()
+	assert.NoError(t, a.Check(ctx, "mathias"))
+	assert.NoError(t, a.Check(ctx, "acme"))
+	assert.Error(t, a.Check(ctx, "evil"))
+	assert.Error(t, a.Check(ctx, ""))
+}
+
+// A caller authenticated with their own Gitea PAT (pass-through, gitea-mcp#59)
+// is gated by Gitea's own permission model, not the MCP's static owner list —
+// otherwise a legitimate second user could never touch their own repos.
+func TestAllowlistCheckTrustsPassthroughAuthenticatedCaller(t *testing.T) {
+	a := allowlist.New([]string{"mathias"})
+	ctx := gitea.WithToken(context.Background(), "someone-elses-pat")
+	assert.NoError(t, a.Check(ctx, "someone-else"))
+}
+
+// Empty owner is a structural input error, not an authz question — still
+// rejected even on the pass-through path.
+func TestAllowlistCheckStillRejectsEmptyOwnerOnPassthrough(t *testing.T) {
+	a := allowlist.New([]string{"mathias"})
+	ctx := gitea.WithToken(context.Background(), "someone-elses-pat")
+	assert.Error(t, a.Check(ctx, ""))
 }
@@ -0,0 +1,100 @@
+// Package localtools holds MCP tools that must never be reachable from the
+// shared HTTP gitea-mcp server (RegisterAll, cmd/gitea-mcp). Each tool here
+// carries a capability too dangerous to expose network-side, even behind
+// OIDC — mint_service_account_token resets whatever password its target
+// account has. Only cmd/gitea-admin-mcp, a separate local-only stdio binary,
+// imports this package.
+package localtools
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"strings"
+
+	"git.d-ma.be/mathias/gitea-mcp/internal/adminbootstrap"
+	"git.d-ma.be/mathias/gitea-mcp/internal/registry"
+)
+
+// bootstrapper is the one method this tool needs — narrow so tests inject a
+// fake instead of a real adminbootstrap.Bootstrapper.
+type bootstrapper interface {
+	Run(ctx context.Context, req adminbootstrap.Request) (adminbootstrap.Result, error)
+}
+
+type MintServiceAccountToken struct {
+	b bootstrapper
+}
+
+func NewMintServiceAccountToken(b bootstrapper) *MintServiceAccountToken {
+	return &MintServiceAccountToken{b: b}
+}
+
+func (t *MintServiceAccountToken) Descriptor() registry.ToolDescriptor {
+	return registry.ToolDescriptor{
+		Name: "mint_service_account_token",
+		Description: "Create (or reset the password of) a dedicated Gitea " +
+			"service/bot account, mint it a scoped access token, optionally " +
+			"grant it read-only collaborator access on named repos, and store " +
+			"the token via 1Password. Requires confirm_service_account=true: " +
+			"this resets whatever password the target account currently has, " +
+			"which is only safe for a dedicated bot account, never a real " +
+			"person's login.",
+		InputSchema: json.RawMessage(`{
+			"type":"object",
+			"properties":{
+				"username":{"type":"string"},
+				"email":{"type":"string"},
+				"full_name":{"type":"string"},
+				"scopes":{"type":"array","items":{"type":"string"}},
+				"token_name":{"type":"string"},
+				"secret_name":{"type":"string","description":"name to store the minted token under"},
+				"grant_read_repos":{"type":"array","items":{"type":"string"},"description":"owner/repo strings"},
+				"confirm_service_account":{"type":"boolean"}
+			},
+			"required":["username","email","scopes","token_name","secret_name","confirm_service_account"]
+		}`),
+	}
+}
+
+type mintArgs struct {
+	Username              string   `json:"username"`
+	Email                 string   `json:"email"`
+	FullName              string   `json:"full_name"`
+	Scopes                []string `json:"scopes"`
+	TokenName             string   `json:"token_name"`
+	SecretName            string   `json:"secret_name"`
+	GrantReadRepos        []string `json:"grant_read_repos"`
+	ConfirmServiceAccount bool     `json:"confirm_service_account"`
+}
+
+func (t *MintServiceAccountToken) Call(ctx context.Context, raw json.RawMessage) (json.RawMessage, error) {
+	var args mintArgs
+	if err := json.Unmarshal(raw, &args); err != nil {
+		return nil, fmt.Errorf("parse arguments: %w", err)
+	}
+
+	repos := make([]adminbootstrap.RepoRef, 0, len(args.GrantReadRepos))
+	for _, r := range args.GrantReadRepos {
+		owner, name, ok := strings.Cut(r, "/")
+		if !ok || owner == "" || name == "" {
+			return nil, fmt.Errorf("grant_read_repos entry %q must be owner/repo", r)
+		}
+		repos = append(repos, adminbootstrap.RepoRef{Owner: owner, Name: name})
+	}
+
+	result, err := t.b.Run(ctx, adminbootstrap.Request{
+		Username:              args.Username,
+		Email:                 args.Email,
+		FullName:              args.FullName,
+		Scopes:                args.Scopes,
+		TokenName:             args.TokenName,
+		SecretName:            args.SecretName,
+		GrantReadRepos:        repos,
+		ConfirmServiceAccount: args.ConfirmServiceAccount,
+	})
+	if err != nil {
+		return nil, err
+	}
+	return json.Marshal(result)
+}
@@ -0,0 +1,110 @@
+package localtools_test
+
+import (
+	"context"
+	"encoding/json"
+	"testing"
+
+	"git.d-ma.be/mathias/gitea-mcp/internal/adminbootstrap"
+	"git.d-ma.be/mathias/gitea-mcp/internal/localtools"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// This tool lives in its own package, never internal/tools, on purpose: it
+// resets whatever password its target account has, which is far too
+// dangerous a capability to risk it ever being wired into RegisterAll and
+// exposed on the shared HTTP gitea-mcp server. Only cmd/gitea-admin-mcp (a
+// separate, local-only, stdio binary) imports this package.
+
+type fakeBootstrapper struct {
+	gotReq adminbootstrap.Request
+	result adminbootstrap.Result
+	err    error
+}
+
+func (f *fakeBootstrapper) Run(_ context.Context, req adminbootstrap.Request) (adminbootstrap.Result, error) {
+	f.gotReq = req
+	return f.result, f.err
+}
+
+func TestMintServiceAccountToken_ParsesArgsAndCallsBootstrapper(t *testing.T) {
+	fb := &fakeBootstrapper{result: adminbootstrap.Result{
+		Username: "dispatch-bot", Created: true,
+		SecretRef: "op://AgentSecrets/gitea-dispatch-bot-ci-readonly/token",
+	}}
+	tool := localtools.NewMintServiceAccountToken(fb)
+
+	args, _ := json.Marshal(map[string]any{
+		"username":                "dispatch-bot",
+		"email":                   "dispatch@d-ma.be",
+		"scopes":                  []string{"read:repository", "read:issue"},
+		"token_name":              "dispatch-ci-readonly",
+		"secret_name":             "gitea-dispatch-bot-ci-readonly",
+		"grant_read_repos":        []string{"mathias/cobalt-dingo"},
+		"confirm_service_account": true,
+	})
+
+	out, err := tool.Call(context.Background(), args)
+
+	require.NoError(t, err)
+	assert.Equal(t, "dispatch-bot", fb.gotReq.Username)
+	assert.Equal(t, []string{"read:repository", "read:issue"}, fb.gotReq.Scopes)
+	assert.Equal(t, true, fb.gotReq.ConfirmServiceAccount)
+	require.Equal(t, []adminbootstrap.RepoRef{{Owner: "mathias", Name: "cobalt-dingo"}}, fb.gotReq.GrantReadRepos)
+
+	var result map[string]any
+	require.NoError(t, json.Unmarshal(out, &result))
+	assert.Equal(t, "op://AgentSecrets/gitea-dispatch-bot-ci-readonly/token", result["SecretRef"])
+	// The tool's own output must never carry a raw token or password field --
+	// adminbootstrap.Result structurally can't (it has none), which this pins.
+	_, hasToken := result["Token"]
+	_, hasPassword := result["Password"]
+	assert.False(t, hasToken)
+	assert.False(t, hasPassword)
+}
+
+func TestMintServiceAccountToken_RejectsMalformedRepoRef(t *testing.T) {
+	fb := &fakeBootstrapper{}
+	tool := localtools.NewMintServiceAccountToken(fb)
+
+	args, _ := json.Marshal(map[string]any{
+		"username":                "dispatch-bot",
+		"email":                   "dispatch@d-ma.be",
+		"scopes":                  []string{"read:repository"},
+		"token_name":              "x",
+		"secret_name":             "x",
+		"grant_read_repos":        []string{"not-owner-slash-repo"},
+		"confirm_service_account": true,
+	})
+
+	_, err := tool.Call(context.Background(), args)
+
+	require.Error(t, err)
+	assert.Contains(t, err.Error(), "owner/repo")
+}
+
+func TestMintServiceAccountToken_PropagatesBootstrapperError(t *testing.T) {
+	fb := &fakeBootstrapper{err: assert.AnError}
+	tool := localtools.NewMintServiceAccountToken(fb)
+
+	args, _ := json.Marshal(map[string]any{
+		"username":                "dispatch-bot",
+		"email":                   "dispatch@d-ma.be",
+		"scopes":                  []string{"read:repository"},
+		"token_name":              "x",
+		"secret_name":             "x",
+		"confirm_service_account": true,
+	})
+
+	_, err := tool.Call(context.Background(), args)
+
+	require.Error(t, err)
+}
+
+func TestMintServiceAccountToken_Descriptor(t *testing.T) {
+	tool := localtools.NewMintServiceAccountToken(&fakeBootstrapper{})
+	d := tool.Descriptor()
+	assert.Equal(t, "mint_service_account_token", d.Name)
+	assert.NotEmpty(t, d.Description)
+}
@@ -0,0 +1,50 @@
+package localtools
+
+import (
+	"context"
+	"fmt"
+	"os/exec"
+	"strings"
+)
+
+// OpSecretStore implements adminbootstrap.SecretStore against the local
+// `op` (1Password) CLI. Run is injectable for tests; nil defaults to
+// actually invoking `op`.
+//
+// The token value is passed via the child process's stdin, never as a
+// literal argv element — matching the estate's own secret-handling rule
+// (a value in argv lands in process listings and any transcript of the
+// command). `credential[password]=-` is `op item create`'s documented way
+// to read one field's value from stdin instead of the command line.
+type OpSecretStore struct {
+	Vault string
+	Run   func(ctx context.Context, stdin string, args ...string) error
+}
+
+func (s OpSecretStore) Store(ctx context.Context, name, value string) (string, error) {
+	run := s.Run
+	if run == nil {
+		run = runOp
+	}
+	args := []string{
+		"item", "create",
+		"--category=login",
+		"--vault=" + s.Vault,
+		"--title=" + name,
+		"credential[password]=-",
+	}
+	if err := run(ctx, value, args...); err != nil {
+		return "", fmt.Errorf("op item create: %w", err)
+	}
+	return fmt.Sprintf("op://%s/%s/credential", s.Vault, name), nil
+}
+
+func runOp(ctx context.Context, stdin string, args ...string) error {
+	cmd := exec.CommandContext(ctx, "op", args...)
+	cmd.Stdin = strings.NewReader(stdin)
+	out, err := cmd.CombinedOutput()
+	if err != nil {
+		return fmt.Errorf("%w: %s", err, strings.TrimSpace(string(out)))
+	}
+	return nil
+}
@@ -0,0 +1,52 @@
+package localtools_test
+
+import (
+	"context"
+	"testing"
+
+	"git.d-ma.be/mathias/gitea-mcp/internal/localtools"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// The estate's own secret-handling rule: never place a literal secret in a
+// command's argv — it lands in process listings and any transcript of the
+// command. OpSecretStore feeds the token to `op` via stdin instead.
+
+func TestOpSecretStore_PassesValueViaStdinNeverArgv(t *testing.T) {
+	var gotStdin string
+	var gotArgs []string
+	s := localtools.OpSecretStore{
+		Vault: "AgentSecrets",
+		Run: func(_ context.Context, stdin string, args ...string) error {
+			gotStdin = stdin
+			gotArgs = args
+			return nil
+		},
+	}
+
+	ref, err := s.Store(context.Background(), "gitea-dispatch-bot-ci-readonly", "tok_the_actual_secret_value")
+
+	require.NoError(t, err)
+	assert.Equal(t, "op://AgentSecrets/gitea-dispatch-bot-ci-readonly/credential", ref)
+	assert.Equal(t, "tok_the_actual_secret_value", gotStdin)
+	for _, a := range gotArgs {
+		assert.NotContains(t, a, "tok_the_actual_secret_value",
+			"the secret value must never appear as a literal argv element")
+	}
+	assert.Contains(t, gotArgs, "--vault=AgentSecrets")
+	assert.Contains(t, gotArgs, "--title=gitea-dispatch-bot-ci-readonly")
+}
+
+func TestOpSecretStore_PropagatesRunError(t *testing.T) {
+	s := localtools.OpSecretStore{
+		Vault: "AgentSecrets",
+		Run: func(_ context.Context, _ string, _ ...string) error {
+			return assert.AnError
+		},
+	}
+
+	_, err := s.Store(context.Background(), "x", "y")
+
+	require.Error(t, err)
+}
@@ -0,0 +1,95 @@
+package mcp
+
+import (
+	"bufio"
+	"context"
+	"encoding/json"
+	"errors"
+	"fmt"
+	"io"
+
+	"git.d-ma.be/mathias/gitea-mcp/internal/registry"
+)
+
+// ServeStdio runs the MCP JSON-RPC protocol over stdio: one JSON-RPC message
+// per line on in, one response per line on out — the standard MCP stdio
+// framing, and how Claude Code spawns a LOCAL MCP server as a subprocess.
+//
+// Built for gitea-admin-mcp (see cmd/gitea-admin-mcp), which has to run
+// locally where real admin credentials exist just-in-time, rather than
+// through the hosted HTTP server (server.go) a sandboxed agent session talks
+// to. The method dispatch below intentionally mirrors handlePOST in
+// server.go — same three methods, same error codes — so a tool written
+// against registry.Tool behaves identically over either transport; it is
+// kept as a separate small loop rather than a shared refactor of the HTTP
+// handler, which mixes in http.ResponseWriter specifics this transport
+// doesn't have.
+func ServeStdio(ctx context.Context, reg *registry.Registry, in io.Reader, out io.Writer) error {
+	scanner := bufio.NewScanner(in)
+	scanner.Buffer(make([]byte, 0, 64*1024), 1<<20) // 1 MiB max line, matches server.go's request cap
+	enc := json.NewEncoder(out)
+
+	for scanner.Scan() {
+		line := scanner.Bytes()
+		if len(line) == 0 {
+			continue
+		}
+
+		var req Request
+		if err := json.Unmarshal(line, &req); err != nil {
+			if werr := enc.Encode(NewErrorResponse(nil, -32700, "parse error", nil)); werr != nil {
+				return werr
+			}
+			continue
+		}
+		if req.ID == nil {
+			continue // notification — no response, per JSON-RPC 2.0
+		}
+
+		resp := dispatch(ctx, reg, req)
+		if err := enc.Encode(resp); err != nil {
+			return err
+		}
+	}
+	return scanner.Err()
+}
+
+// dispatch handles one already-parsed request. Mirrors server.go's
+// handlePOST switch exactly (initialize / tools/list / tools/call / unknown).
+func dispatch(ctx context.Context, reg *registry.Registry, req Request) Response {
+	if req.Method == "initialize" {
+		return NewResponse(req.ID, map[string]any{
+			"protocolVersion": ProtocolVersion,
+			"capabilities":    map[string]any{"tools": map[string]any{}},
+			"serverInfo":      map[string]any{"name": "gitea-admin-mcp", "version": "0.1.0"},
+		})
+	}
+
+	switch req.Method {
+	case "tools/list":
+		return NewResponse(req.ID, map[string]any{"tools": reg.Tools()})
+
+	case "tools/call":
+		var p struct {
+			Name      string          `json:"name"`
+			Arguments json.RawMessage `json:"arguments"`
+		}
+		if err := json.Unmarshal(req.Params, &p); err != nil {
+			return NewErrorResponse(req.ID, -32602, "invalid params", nil)
+		}
+		out, err := reg.Dispatch(ctx, p.Name, p.Arguments)
+		if err != nil {
+			code := -32000
+			if errors.Is(err, registry.ErrToolNotFound) {
+				code = CodeNotFound
+			}
+			return NewErrorResponse(req.ID, code, err.Error(), nil)
+		}
+		return NewResponse(req.ID, map[string]any{
+			"content": []map[string]any{{"type": "text", "text": string(out)}},
+		})
+
+	default:
+		return NewErrorResponse(req.ID, -32601, fmt.Sprintf("method not found: %s", req.Method), nil)
+	}
+}
@@ -0,0 +1,115 @@
+package mcp_test
+
+import (
+	"bytes"
+	"context"
+	"encoding/json"
+	"strings"
+	"testing"
+
+	"git.d-ma.be/mathias/gitea-mcp/internal/mcp"
+	"git.d-ma.be/mathias/gitea-mcp/internal/registry"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// The existing MCP server (server.go) only speaks HTTP. gitea-admin-mcp needs
+// to run as a LOCAL stdio server (Claude Code's local .mcp.json stdio server
+// type, spawned as a subprocess with real 1Password/admin auth already in its
+// environment) -- reusing the exact same registry.Tool interface and
+// jsonrpc.go message types, one JSON-RPC message per line on stdin/stdout,
+// which is the standard MCP stdio framing.
+
+type echoTool struct{}
+
+func (echoTool) Descriptor() registry.ToolDescriptor {
+	return registry.ToolDescriptor{Name: "echo", Description: "echoes its input"}
+}
+func (echoTool) Call(_ context.Context, args json.RawMessage) (json.RawMessage, error) {
+	return args, nil
+}
+
+func TestServeStdio_Initialize(t *testing.T) {
+	reg := registry.New()
+	in := strings.NewReader(`{"jsonrpc":"2.0","id":1,"method":"initialize","params":{}}` + "\n")
+	out := &bytes.Buffer{}
+
+	err := mcp.ServeStdio(context.Background(), reg, in, out)
+
+	require.NoError(t, err)
+	var resp mcp.Response
+	require.NoError(t, json.Unmarshal(bytes.TrimSpace(out.Bytes()), &resp))
+	assert.EqualValues(t, 1, resp.ID)
+	assert.Nil(t, resp.Error)
+}
+
+func TestServeStdio_ToolsListAndCall(t *testing.T) {
+	reg := registry.New()
+	reg.Register(echoTool{})
+	in := strings.NewReader(
+		`{"jsonrpc":"2.0","id":1,"method":"tools/list"}` + "\n" +
+			`{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"echo","arguments":{"hello":"world"}}}` + "\n",
+	)
+	out := &bytes.Buffer{}
+
+	err := mcp.ServeStdio(context.Background(), reg, in, out)
+	require.NoError(t, err)
+
+	lines := strings.Split(strings.TrimSpace(out.String()), "\n")
+	require.Len(t, lines, 2, "one JSON-RPC response per line, in request order")
+
+	var listResp mcp.Response
+	require.NoError(t, json.Unmarshal([]byte(lines[0]), &listResp))
+	assert.EqualValues(t, 1, listResp.ID)
+
+	var callResp mcp.Response
+	require.NoError(t, json.Unmarshal([]byte(lines[1]), &callResp))
+	assert.EqualValues(t, 2, callResp.ID)
+	require.NotNil(t, callResp.Result)
+}
+
+func TestServeStdio_NotificationGetsNoResponse(t *testing.T) {
+	// A request with no "id" is a notification per JSON-RPC 2.0 -- must not
+	// write anything back, and must not stop the loop from processing what
+	// follows it.
+	reg := registry.New()
+	in := strings.NewReader(
+		`{"jsonrpc":"2.0","method":"notifications/initialized"}` + "\n" +
+			`{"jsonrpc":"2.0","id":1,"method":"tools/list"}` + "\n",
+	)
+	out := &bytes.Buffer{}
+
+	err := mcp.ServeStdio(context.Background(), reg, in, out)
+	require.NoError(t, err)
+
+	lines := strings.Split(strings.TrimSpace(out.String()), "\n")
+	require.Len(t, lines, 1, "the notification must not produce a response line")
+}
+
+func TestServeStdio_UnknownMethodReturnsJSONRPCError(t *testing.T) {
+	reg := registry.New()
+	in := strings.NewReader(`{"jsonrpc":"2.0","id":1,"method":"not/a/real/method"}` + "\n")
+	out := &bytes.Buffer{}
+
+	err := mcp.ServeStdio(context.Background(), reg, in, out)
+	require.NoError(t, err)
+
+	var resp mcp.Response
+	require.NoError(t, json.Unmarshal(bytes.TrimSpace(out.Bytes()), &resp))
+	require.NotNil(t, resp.Error)
+	assert.Equal(t, -32601, resp.Error.Code)
+}
+
+func TestServeStdio_UnparseableLineReturnsParseError(t *testing.T) {
+	reg := registry.New()
+	in := strings.NewReader("not json at all\n")
+	out := &bytes.Buffer{}
+
+	err := mcp.ServeStdio(context.Background(), reg, in, out)
+	require.NoError(t, err, "a malformed line must not crash the loop")
+
+	var resp mcp.Response
+	require.NoError(t, json.Unmarshal(bytes.TrimSpace(out.Bytes()), &resp))
+	require.NotNil(t, resp.Error)
+	assert.Equal(t, -32700, resp.Error.Code)
+}
@@ -46,7 +46,7 @@ func (t *BranchDelete) Call(ctx context.Context, raw json.RawMessage) (json.RawM
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	if args.Branch == "" {
@@ -47,7 +47,7 @@ func (t *BranchList) Call(ctx context.Context, raw json.RawMessage) (json.RawMes
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}

@@ -45,7 +45,7 @@ func (t *BranchProtectionGet) Call(ctx context.Context, raw json.RawMessage) (js
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}

@@ -70,7 +70,7 @@ func (t *CodeSearch) Call(ctx context.Context, raw json.RawMessage) (json.RawMes
 	if args.Q == "" {
 		return nil, fmt.Errorf("q is required: %w", gitea.ErrValidation)
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	if args.Page < 1 {
@@ -111,7 +111,7 @@ func (t *CreateProjectFromTemplate) Call(ctx context.Context, raw json.RawMessag
 	}

 	// Allowlist check first.
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}

@@ -47,7 +47,7 @@ func (t *DirList) Call(ctx context.Context, raw json.RawMessage) (json.RawMessag
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}

@@ -52,7 +52,7 @@ func (t *FileDelete) Call(ctx context.Context, raw json.RawMessage) (json.RawMes
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	if args.Sha == "" {
@@ -51,7 +51,7 @@ func (t *FileRead) Call(ctx context.Context, raw json.RawMessage) (json.RawMessa
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}

@@ -57,7 +57,7 @@ func (t *FileWriteBranch) Call(ctx context.Context, raw json.RawMessage) (json.R
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	if args.Branch == "" {
@@ -45,7 +45,7 @@ func (t *IssueClose) Call(ctx context.Context, raw json.RawMessage) (json.RawMes
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	iss, err := t.c.SetIssueState(ctx, args.Owner, args.Repo, args.Number, "closed")
@@ -50,7 +50,7 @@ func (t *IssueComment) Call(ctx context.Context, raw json.RawMessage) (json.RawM
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	if args.Number < 1 {
@@ -56,7 +56,7 @@ func (t *IssueCreate) Call(ctx context.Context, raw json.RawMessage) (json.RawMe
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	if args.Title == "" {
@@ -53,7 +53,7 @@ func (t *IssueEdit) Call(ctx context.Context, raw json.RawMessage) (json.RawMess
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	if args.Number < 1 {
@@ -43,7 +43,7 @@ func (t *IssueGet) Call(ctx context.Context, raw json.RawMessage) (json.RawMessa
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	iss, err := t.c.GetIssue(ctx, args.Owner, args.Repo, args.Number)
@@ -50,7 +50,7 @@ func (t *IssueLabel) Call(ctx context.Context, raw json.RawMessage) (json.RawMes
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	if args.Number < 1 {
@@ -53,7 +53,7 @@ func (t *IssueList) Call(ctx context.Context, raw json.RawMessage) (json.RawMess
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	if args.State == "" {
@@ -45,7 +45,7 @@ func (t *IssueListComments) Call(ctx context.Context, raw json.RawMessage) (json
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	comments, err := t.c.ListIssueComments(ctx, args.Owner, args.Repo, args.Number)
@@ -45,7 +45,7 @@ func (t *IssueReopen) Call(ctx context.Context, raw json.RawMessage) (json.RawMe
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	iss, err := t.c.SetIssueState(ctx, args.Owner, args.Repo, args.Number, "open")
@@ -43,7 +43,7 @@ func (t *LabelList) Call(ctx context.Context, raw json.RawMessage) (json.RawMess
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	labels, err := t.c.ListLabels(ctx, args.Owner, args.Repo)
@@ -50,7 +50,7 @@ func (t *PRComment) Call(ctx context.Context, raw json.RawMessage) (json.RawMess
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	if args.Number < 1 {
@@ -56,7 +56,7 @@ func (t *PRCreate) Call(ctx context.Context, raw json.RawMessage) (json.RawMessa
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	if args.Title == "" {
@@ -63,7 +63,7 @@ func (t *PRFilesDiff) Call(ctx context.Context, raw json.RawMessage) (json.RawMe
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	if args.Number < 1 {
@@ -44,7 +44,7 @@ func (t *PRGet) Call(ctx context.Context, raw json.RawMessage) (json.RawMessage,
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	if args.Number < 1 {
@@ -51,7 +51,7 @@ func (t *PRList) Call(ctx context.Context, raw json.RawMessage) (json.RawMessage
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	state := args.State
@@ -52,7 +52,7 @@ func (t *PRMerge) Call(ctx context.Context, raw json.RawMessage) (json.RawMessag
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	if args.Number < 1 {
@@ -55,7 +55,7 @@ func (t *ReleaseCreate) Call(ctx context.Context, raw json.RawMessage) (json.Raw
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	rel, err := t.c.CreateRelease(ctx, args.Owner, args.Repo, gitea.CreateReleaseArgs{
@@ -53,7 +53,7 @@ func (t *RepoCreate) Call(ctx context.Context, raw json.RawMessage) (json.RawMes
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	createArgs := gitea.CreateRepoArgs{
@@ -46,7 +46,7 @@ func (t *RepoDelete) Call(ctx context.Context, raw json.RawMessage) (json.RawMes
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	if args.Confirm != args.Repo {
@@ -38,7 +38,7 @@ func (t *RepoGet) Call(ctx context.Context, raw json.RawMessage) (json.RawMessag
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	r, err := t.c.GetRepo(ctx, args.Owner, args.Repo)
@@ -45,7 +45,7 @@ func (t *RepoList) Call(ctx context.Context, raw json.RawMessage) (json.RawMessa
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	args.Limit = capLimit(args.Limit, 30)
@@ -96,7 +96,7 @@ func (t *RepoMirrorPush) Call(ctx context.Context, raw json.RawMessage) (json.Ra
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	switch args.Action {
@@ -53,7 +53,7 @@ func (t *RepoSearch) Call(ctx context.Context, raw json.RawMessage) (json.RawMes
 		return nil, fmt.Errorf("q is required: %w", gitea.ErrValidation)
 	}
 	if args.Owner != "" {
-		if err := t.a.Check(args.Owner); err != nil {
+		if err := t.a.Check(ctx, args.Owner); err != nil {
 			return nil, err
 		}
 	}
@@ -75,7 +75,7 @@ func (t *RepoSearch) Call(ctx context.Context, raw json.RawMessage) (json.RawMes
 			if len(parts) != 2 {
 				continue
 			}
-			if t.a.Check(parts[0]) == nil {
+			if t.a.Check(ctx, parts[0]) == nil {
 				filtered = append(filtered, r)
 			}
 		}
@@ -45,7 +45,7 @@ func (t *RepoStatus) Call(ctx context.Context, raw json.RawMessage) (json.RawMes
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}

@@ -45,7 +45,7 @@ func (t *RepoTopicsUpdate) Call(ctx context.Context, raw json.RawMessage) (json.
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	if err := t.c.UpdateTopics(ctx, args.Owner, args.Repo, args.Topics); err != nil {
@@ -45,7 +45,7 @@ func (t *RepoTree) Call(ctx context.Context, raw json.RawMessage) (json.RawMessa
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	tree, err := t.c.GetTree(ctx, args.Owner, args.Repo, args.Ref, true)
@@ -60,7 +60,7 @@ func (t *RepoUpdate) Call(ctx context.Context, raw json.RawMessage) (json.RawMes
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}

@@ -50,7 +50,7 @@ func (t *TagCreate) Call(ctx context.Context, raw json.RawMessage) (json.RawMess
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	if args.Tag == "" {
@@ -145,7 +145,7 @@ func (t *TBDShip) Call(ctx context.Context, raw json.RawMessage) (json.RawMessag
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	if args.Path == "" || args.Content == "" || args.Message == "" {
@@ -57,7 +57,7 @@ func (t *WorkflowRunList) Call(ctx context.Context, raw json.RawMessage) (json.R
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	args.Limit = capLimit(args.Limit, 10)
@@ -47,7 +47,7 @@ func (t *WorkflowRunStatus) Call(ctx context.Context, raw json.RawMessage) (json
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	if args.RunID < 1 {
@@ -62,7 +62,7 @@ func (t *WorkflowRunTrigger) Call(ctx context.Context, raw json.RawMessage) (jso
 	if err := parseArgs(raw, &args); err != nil {
 		return nil, err
 	}
-	if err := t.a.Check(args.Owner); err != nil {
+	if err := t.a.Check(ctx, args.Owner); err != nil {
 		return nil, err
 	}
 	if args.Workflow == "" {
@@ -5,6 +5,11 @@
 # For Claude Code: generates project-only CLAUDE.md (it inherits root via tree walk)
 # For everything else: concatenates root AGENT.md + project PROJECT.md
 #
+# Cursor and Aider adapters removed 2026-09-01 — neither is a supported harness
+# any more. Supported: Claude Code, Antigravity, OpenCode, Mistral Vibe, Crush,
+# Pi. All read CLAUDE.md or AGENTS.md; system-prompt.txt remains for Open WebUI,
+# Mods and raw API use.
+#
 # Usage: ./scripts/context-sync.sh [--force] [adapter...]
 # Task:  task context:sync
 #
@@ -95,30 +100,6 @@ generate_agents() {
  echo "  → AGENTS.md (root + project; Crush, Pi, Antigravity)"
 }

-# ── Cursor ───────────────────────────────────────────────────
-generate_cursor() {
-  {
-    echo "# Cursor rules — auto-generated"
-    echo "# Do not edit. Run: task context:sync"
-    echo ""
-    root_block
-    cat "$PROJECT_FILE"
-  } > .cursorrules
-  echo "  → .cursorrules (root + project)"
-}
-
-# ── Aider ────────────────────────────────────────────────────
-generate_aider() {
-  { root_block; cat "$PROJECT_FILE"; } > .aider.conventions.md
-  if [ ! -f .aider.conf.yml ]; then
-    cat > .aider.conf.yml << 'YAML'
-read: .aider.conventions.md
-auto-commits: false
-YAML
-  fi
-  echo "  → .aider.conventions.md (root + project)"
-}
-
 # ── Generic system prompt (Open WebUI, Mods, etc.) ──────────
 generate_system_prompt() {
  {
@@ -180,8 +161,6 @@ echo "Syncing project context from $PROJECT_FILE..."
 if [ ${#ADAPTERS[@]} -eq 0 ]; then
  generate_claude
  generate_agents
-  generate_cursor
-  generate_aider
  generate_system_prompt
  generate_mcp
 else
@@ -189,8 +168,6 @@ else
    case "$adapter" in
      claude)  generate_claude ;;
      agents)  generate_agents ;;
-      cursor)  generate_cursor ;;
-      aider)   generate_aider ;;
      prompt|system|openwebui|owui|generic) generate_system_prompt ;;
      mcp)     generate_mcp ;;
      *) echo "Unknown adapter: $adapter" >&2; exit 1 ;;
Author	SHA1	Message	Date
mathias	ef7ccc8295	fix(lint): check resp.Body.Close()'s error (errcheck) CD / Deploy via GitOps (push) Successful in 5s Details CD / Lint / Test / Vet (push) Successful in 18s Details CD / Build & Import (push) Successful in 6s Details Caught by CI, not locally -- this sandbox's golangci-lint crashes on any repo, a pre-existing environment issue unrelated to this code (documented in dispatch's own session history). gitea-mcp's actual CI run has a working linter and correctly flagged this.	2026-09-05 11:27:22 +02:00
mathias	4d6597639e	feat(admin): local-only stdio MCP server to mint scoped service-account tokens (dispatch#67) CD / Lint / Test / Vet (push) Failing after 4s Details CD / Build & Import (push) Skipped Details CD / Deploy via GitOps (push) Skipped Details dispatch#67 fix #2 needed a read-only CI credential and hit a hard wall: Gitea's own token-creation and admin-user endpoints reject token auth outright -- POST /users/{u}/tokens returns 'auth required', POST /admin/users returns 'token does not have at least one of required scope(s), required=[write:admin]' -- for ANY existing token, by design. An existing token cannot mint another token or create an account; the only way through either endpoint is real basic auth as an actual site admin. That is exactly what a sandboxed agent session cannot safely hold. internal/adminbootstrap: the core bootstrap, HTTP-only, fully testable against a fake transport (5 tests, no network). Given real admin username/password, it creates the target account (or resets its password if it already exists -- deliberately unconditional, gated behind a required ConfirmServiceAccount flag since that reset is only safe for a dedicated bot account), mints it a scoped token by authenticating AS that account (Gitea requires this specifically -- an admin cannot mint on another account's behalf), grants it read-only collaborator access on named repos, and stores the result via a SecretStore -- never returning the token or either password to the caller. internal/localtools: the MCP tool wrapper (mint_service_account_token) and the real op-CLI-backed SecretStore (token passed via stdin, never argv, per the estate's own secret-handling rule). Deliberately its OWN package, not internal/tools -- this capability is far too dangerous to risk it ever being wired into RegisterAll and reachable from the shared HTTP gitea-mcp server, even behind OIDC. internal/mcp/stdio.go: a stdio JSON-RPC transport (one message per line, standard MCP stdio framing) for the existing registry.Tool interface -- the only transport gitea-mcp had before this was HTTP (server.go), which a sandboxed session can reach but a real admin password never should be piped through. cmd/gitea-admin-mcp: the binary. Meant to be spawned by a LOCAL Claude Code session (.mcp.json stdio server) or run by hand, via 'op run --env-file ~/.op-env' so admin credentials arrive just-in-time and never touch a sandboxed session. 11 new tests total, all passing under -race. Nothing in cmd/gitea-mcp (the shared HTTP server) changed.	2026-09-05 11:25:41 +02:00
mathias	419ae19cdb	fix(context): sync to the estate canonical context-sync.sh, drop Cursor/Aider CD / Lint / Test / Vet (push) Successful in 24s Details CD / Build & Import (push) Successful in 14s Details CD / Deploy via GitOps (push) Successful in 5s Details This repo's copy of context-sync.sh was two days stale against the version Mathias's Cursor/Aider decision (2026-09-01) already landed in dma-cli: still generated .cursorrules and .aider.conventions.md, and lacked the --force/placeholder pre-flight and .context/mcp.json root-merge dma-cli's copy has had since. Found while building the estate-wide context-sync drift gate (ADR-0019) -- this is exactly the drift the gate exists to catch, found by hand while establishing the baseline it will check going forward. - script synced to the canonical hash (dma-cli's current copy) - .cursorrules, .aider.conventions.md, .aider.conf.yml deleted - Taskfile's drift assertion and file list narrowed to match - AGENTS.md, .context/system-prompt.txt regenerated against canonical task check -> exit 0.	2026-09-05 08:01:07 +02:00
mathiasandClaude Opus 5	e4d07c106c	ci(deploy): refuse to patch a manifest with an empty IMAGE_TAG (infra#311) CD / Lint / Test / Vet (push) Successful in 10s Details CD / Build & Import (push) Successful in 7s Details CD / Deploy via GitOps (push) Successful in 9s Details On 2026-08-25 a deploy job ran with an empty IMAGE_TAG, wrote `image: localhost:5000/gitea-mcp:` into the infra repo and pushed it. That is invalid YAML, so Flux failed the ENTIRE apps Kustomization -- every app stopped reconciling for about an hour, including three unrelated deploys queued behind it. This job's existing check could not catch it. It greps for `image: <repo>:${IMAGE_TAG}` after the sed, and with IMAGE_TAG empty that pattern matches the corrupted line itself, so the verification passes because of the bug rather than despite it. Asserting the input up front is the fix: an empty tag now fails the job before anything is written, let alone committed. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_011PL5XcG68SvQqz3vGVbWbf	2026-08-25 17:25:23 +02:00
mathiasandClaude Opus 5	1d75bf9b25	fix(ci): no expression syntax in a run: block's comments CD / Lint / Test / Vet (push) Successful in 10s Details CD / Deploy via GitOps (push) Successful in 4s Details CD / Build & Import (push) Successful in 5s Details act interpolates a `run:` block INCLUDING its shell comments, so prose containing Actions expression syntax is evaluated as an expression. A comment I added yesterday explaining why a fallback was computed in shell contained that syntax as an example; act tried to evaluate it, failed with "Unavailable context: a", and the step then wrote NO outputs at all. In gitea-mcp that emptied `needs.build.outputs.image-tag`, so its deploy job committed `image: localhost:5000/gitea-mcp:` to the infra repo -- which is not valid YAML, so Flux's apps Kustomization went Ready:False and NOTHING in k3s/apps reconciled estate-wide until infra 51ceb7c. The build still went green throughout, because the shared build-image.sh falls back to `git rev-parse --short HEAD` when SHA_TAG is empty. A robust default hid a broken input -- worth remembering as a pattern. Same comment text removed here; the explanation stays, the syntax does not. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_011PL5XcG68SvQqz3vGVbWbf	2026-08-25 15:29:40 +02:00
mathiasandClaude Opus 5	417ab60707	ci: call the shared build-image script instead of an inline copy (infra#309) CD / Lint / Test / Vet (push) Successful in 9s Details CD / Build & Import (push) Successful in 6s Details CD / Deploy via GitOps (push) Failing after 1m2s Details Wave 2. The buildctl/skopeo step here was one of nineteen near-identical copies; the implementation now lives in .gitea/build-image.sh, byte-identical to mathias/infra's ci-templates/build-image.sh and held there by a drift gate. Labels, tags and build-args are inputs. Behaviour unchanged: same registry, same :sha/:latest/version tags, same revision and source labels. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_011PL5XcG68SvQqz3vGVbWbf	2026-08-25 12:35:52 +02:00
mathiasandClaude Opus 5	f6446c2515	ci: drop the k3s ctr smoke test, impossible on the containerized runner CD / Lint / Test / Vet (push) Successful in 8s Details CD / Build & Import (push) Successful in 7s Details CD / Deploy via GitOps (push) Successful in 6s Details The step ran `sudo k3s ctr images pull` / `ctr run` against the host's containerd. That worked when the runner was the Arch host; since it was containerized (infra#233) the image has no sudo, no ctr and no containerd socket, so the step fails with `sudo: command not found`. It only surfaced now because the buildah wall (infra#241) killed the job before this step ever ran. The deploy job's `Verify rollout` and `Confirm pod running new image` assert the same property against the real workload rather than a 5s throwaway container, which is why tapir dropped its smoke test under the same issue. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_011PL5XcG68SvQqz3vGVbWbf	2026-08-25 06:54:55 +02:00
mathiasandClaude Opus 5	50c447f673	ci(cd): build with buildctl instead of rootless buildah (infra#241) CD / Build & Import (push) Failing after 14s Details CD / Deploy via GitOps (push) Skipped Details CD / Lint / Test / Vet (push) Successful in 27s Details Rootless buildah in the containerized gitea-runner pod cannot remap UID/GID on a fresh base-image pull and has no OCI runtime, so this Dockerfile's `RUN apk add` / `RUN go build` cannot execute there at all. The last green run here (2026-07-22) predates the runner's containerization (infra#233, 2026-07-29), so the build has been latently broken since. buildctl hands the build to koala's host buildkitd over the socket the pod already mounts; skopeo pushes the resulting OCI archive once per tag, since buildctl has no `buildah push` equivalent. The VERSION build-arg carries over as `--opt build-arg:VERSION=`, so the ldflags version stamp is unchanged. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_011PL5XcG68SvQqz3vGVbWbf	2026-08-25 06:49:47 +02:00
mathiasandClaude Sonnet 5	43714047be	fix(auth): owner allowlist trusts pass-through-authenticated callers (#59 ) CD / Lint / Test / Vet (push) Successful in 9s Details CD / Build & Import (push) Successful in 25s Details CD / Deploy via GitOps (push) Has been skipped Details Allowlist.Check now takes ctx: when the caller authenticated with their own Gitea PAT (pass-through, v0.12.0), it skips the static GITEA_MCP_ALLOWED_OWNERS check entirely — Gitea's own permission model already gates that caller's access more precisely than a coarse owner-name list can. The static list still applies unchanged for the shared static-token/JWT path, where it's the only defense against the service token's blast radius. Mechanical: every tool call site already had ctx in scope, so this is a signature-only change at 41 call sites, no other tool behavior changes. Closes the "Deferred" item from #59. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>	2026-07-22 08:31:43 +02:00