Template
mathiasandClaude Opus 5
2b3851a09e
Closes #1. Closes #2. A repo generated from this template failed `task check` on errcheck from its first commit, while the template's own `task check` was green. The reason is not in the code: Go ignores directories whose names begin with "_" or "." when expanding package patterns like ./... The entrypoint lives in cmd/__PROJECT_NAME__. So `go build ./...`, `go vet ./...`, `go test ./...` and `golangci-lint run ./...` ALL silently skip the one package every generated repo is built around: $ go list ./... __MODULE_PATH__/internal/agent __MODULE_PATH__/pkg/litellm <- no cmd package After substitution the directory has a real name, becomes visible, and the defect appears. Template green, children red, by construction -- and invisible to anyone reading either repo's CI. Fixes the one live defect: `defer shutdown(ctx)` dropped an error, and SetupTelemetry returns func(context.Context) error. Two of the three defects these issues describe were already fixed and neither issue knew: go.sum has been committed for some time, and resp.Body.Close is already wrapped. Verified before touching anything rather than trusted from the issue text. Adds scripts/tests/generated-child_test.sh, wired into `task check`: it substitutes the tracked tree into a throwaway child and gates go list / build / vet / golangci-lint on the RESULT. Without it the next entrypoint defect rots exactly the same way. Not done by adding ./cmd/__PROJECT_NAME__ to the lint command, which does work -- naming a path directly bypasses the pattern exclusion -- because it reports a gofumpt violation on the import block that does not exist in any child. The formatter's import grouping depends on the module path and the template's is a placeholder, so that route turns the gate red over an artifact of being a template. The harness reads tracked files from the working tree rather than `git archive HEAD`: it runs inside `task check`, so it must gate what is about to be committed. Archiving HEAD reported a stale pass against an unstaged fix on first use here. Proven to fail when it should: red on the unfixed entrypoint, and red again on the stale-HEAD read, before going green. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
PROJECT_NAME
Go agent built on Google ADK with a LiteLLM adapter for local model routing.
Quick start
cp .env.example .env
# edit .env with your LITELLM_API_KEY
go mod tidy
task run
Harness & workflow
Single harness: Claude Code + hyperguild — session-based (start a session, it works, it
stops). No push-triggered dispatch. MCP connections live in .mcp.json: brain (knowledge)
and gitea (issues/PRs — the audit trail); provide BRAIN_MCP_TOKEN and GITEA_MCP_TOKEN.
Work flows: issue → hyperguild session → PR (task check gate) → report-back comment on the
issue → brain capture if the learning is durable. See AGENTS.md / CLAUDE.md for the full
context, AGENT_BOUNDARIES.md for runtime scope.
Observability
Set OTLP_ENDPOINT=http://jaeger.d-ma.be:4318 to emit traces. Each invocation produces:
invoke_agent __PROJECT_NAME__spangenerate_content <model>child span withgen_ai.request.modelattribute
Structure
cmd/__PROJECT_NAME__/ agent entrypoint
pkg/litellm/ OpenAI-compat ADK adapter + OTLP telemetry helper