feat(pipeline): skip extraction when the source has not changed (#105)

Ingest now stamps source_hash — sha256 of the ingested body, frontmatter
excluded — onto the source page, and compares it before doing any work. An
unchanged source returns Skipped with no LLM call at all.

This is cost and signal, not correctness: #103 already made a re-ingest
byte-identical, so a repeated run was harmless but not free — a full
extraction over every chunk to produce a result that merges to nothing.
Building this first would have put idempotency in the guard rather than in
the write, which is the wrong dependency direction.

Body-only hashing on purpose. A re-crawl that stamps a new fetch date has
not changed the source, and treating that as drift would re-extract the
whole corpus every time a crawler ran. A missing or unreadable hash always
means "extract": a missed skip costs one extraction, a wrong skip loses an
update silently, and silence is the failure mode this repo keeps paying
for.

The skip is reported (Result.Skipped plus a warning naming the hash), not
silent — a run that does nothing must say so, or "the pipeline is fine" and
"the pipeline stopped working" look identical (#107).

Storing the hash required settling who owns which frontmatter field. #103
made the on-disk block win wholesale, which protects human-set metadata
from a regenerating writer but froze fields the generator legitimately
owns. wiki.Merge now refreshes an explicit generatorOwned allowlist —
last_updated and source_hash — and keeps everything else on disk. An
allowlist rather than a carve-out, so a field nobody has thought about
stays human-owned by default. That also retires the last_updated
limitation #103 shipped with.

Refs #105, #103, #107, brain#21.

.dispatch-allow

@@ -0,0 +1,2 @@
+# Presence of this file marks this repo as opt-in for headless dispatch.
+# See dispatch#3.

README.md

@@ -5,6 +5,11 @@ Instead of letting Claude Code do whatever it wants, hyperguild enforces structu
 workflows (TDD red/green/refactor), logs every session, and accumulates learnings
 into a searchable brain.
 
+**This repo is also where `brain-mcp` lives** — the MCP server at
+`brain-mcp.d-ma.be` that `brain_query`, `brain_write`, `brain_read`, `brain_edit`
+and `capture` are served from. It is not a separate deployable; see
+[Where brain-mcp lives](#where-brain-mcp-lives).
+
 ## Hypothesis
 
 > We believe routing skill tasks through local models, backed by brain context,
@@ -20,7 +25,7 @@ usage, the fast-model routing path should be reconsidered.
 
 **hyperguild = Claude Code + MCP.** This is a supervisor for Claude Code sessions specifically.
 For multi-agent orchestration (OpenCode + LiteLLM, executor/reviewer pipelines), see
-[agentsquad](http://gitea.d-ma.be/mathias/agentsquad) — a separate harness for a different
+[agentsquad](https://git.d-ma.be/mathias/agentsquad) — a separate harness for a different
 orchestration model. Skills (mathias/skills) are shared between both.
 
 ## How it works
@@ -44,6 +49,73 @@ brain/
 └── knowledge/      — legacy flat notes (migration pending: hyperguild#22)
 ```
 
+## Where brain-mcp lives
+
+`brain-mcp` is served from **`ingestion/internal/mcp`** in this repo, at
+**`brain-mcp.d-ma.be`**. It is a component *inside* the `ingestion` service, not
+a standalone deployable — one pod in the `supervisor` namespace serves the MCP
+endpoint, the REST API, and the claudewatcher together.
+
+| You are looking for | It is here |
+|---|---|
+| The MCP tool surface (`brain_query`, `brain_write`, `brain_read`, `brain_edit`, `capture`, …) | `ingestion/internal/mcp` |
+| The HTTP/REST API (`/query`, `/write`, `/ingest`) | `ingestion/internal/api` |
+| Session ingestion from Claude Code transcripts | `ingestion/internal/claudewatcher` |
+| The brain *content* (notes, wings, halls) | [`mathias/brain`](https://git.d-ma.be/mathias/brain) — a separate repo |
+| Deployment manifests | `config/supervisor` |
+
+### Why it is not its own repo
+
+Asked and decided in [#91](https://git.d-ma.be/mathias/hyperguild/issues/91)
+(part 2). `internal/mcp` imports 17 of ingestion's other internal packages and
+has exactly one importer — it is the MCP facade over the service's own
+internals, so "move it out" resolves to renaming `ingestion`, publishing 17
+packages as a shared library, or maintaining a shim. The discoverability problem
+that prompted the question is answered by naming it here instead.
+
+Full reasoning, including the reversal trigger: brain
+`wiki/hyperguild/decisions/brain-mcp-stays-in-ingestion-fix-discoverability-by-name`.
+
+## Ingest revises, it does not overwrite
+
+Re-ingesting a source **merges into the page already on disk** rather than
+replacing it. Previously the write was unconditional, so a second ingest of the
+same source discarded everything accumulated since — including hand-written
+sections.
+
+The merge is deliberately asymmetric in the existing page's favour:
+
+| On the page | In the fresh extraction | Result |
+|---|---|---|
+| a section | the same section | **the on-disk version wins** |
+| — | a new section | added |
+| a bullet list (`Sources`, `Key Claims`, …) | more bullets | unioned, no duplicates |
+| an append section (`Open Questions`, `Updates`, …) | more lines | only genuinely new lines appended |
+| frontmatter | frontmatter | **the on-disk frontmatter wins** |
+
+So an extraction can add to a note but never silently rewrite what a human put
+there, which is what makes it safe to run on every ingest rather than behind a
+review gate.
+
+Re-ingesting unchanged content is **byte-identical** — no diff, so brain-sync
+does not commit noise that hides real edits.
+
+Two fields are **generator-owned** and do get refreshed on a revise:
+`last_updated` and `source_hash`. Everything else on disk wins. The boundary is
+an explicit allowlist, so a field nobody has thought about stays human-owned.
+
+## Re-ingesting an unchanged source is skipped
+
+Ingest stores `source_hash` — sha256 of the ingested **body**, frontmatter
+excluded — on the source page. A later ingest of the same source compares it
+first and skips extraction entirely when it matches, reporting the skip rather
+than doing it silently.
+
+Body-only is deliberate: a re-crawl that stamps a new fetch date has not changed
+the source, and treating that as drift would re-extract the corpus every time a
+crawler ran. A missing or unreadable hash always means "extract" — a missed skip
+costs one extraction, a wrong skip loses an update.
+
 ## Phase 1 tools (available now)
 
 | Tool | What it does |
@@ -58,7 +130,7 @@ brain/
 | `tier` | Returns the current connectivity tier (1=cloud, 2=LAN, 3=offline) |
 
 > **Note:** `tdd_red/green/refactor` and `spec` were retired in Plan 7 (2026-05-12).
-> They are now SKILL.md files in [mathias/skills](http://gitea.d-ma.be/mathias/skills).
+> They are now SKILL.md files in [mathias/skills](https://git.d-ma.be/mathias/skills).
 
 ## Connect a project
 
@@ -147,7 +219,7 @@ The routing pod selects models per skill call based on historical pass rate:
 
 ## Open issues
 
-See [issues](http://gitea.d-ma.be/mathias/hyperguild/issues) — key open items:
+See [issues](https://git.d-ma.be/mathias/hyperguild/issues) — key open items:
 
 - **#25** — skills platform overhaul (audit first, then lazy loading + brain feedback loop)
 - **#24** — reduce context burn from skill listing

ingestion/cmd/server/main.go

@@ -144,6 +144,16 @@ func envInt(key string, fallback int) int {
 	return fallback
 }
 
+// buildNotifier returns the ntfy notifier, or a nil interface when
+// BRAIN_NTFY_URL is unset. A nil *interface* (not a typed-nil) is required so
+// DegradingSink/Reconcile and the governance observer skip it cleanly.
+func buildNotifier() audit.Notifier {
+	if n := audit.NewNtfyNotifier(os.Getenv("BRAIN_NTFY_URL"), os.Getenv("BRAIN_NTFY_TOKEN")); n != nil {
+		return n
+	}
+	return nil
+}
+
 // buildAuditSink selects the capture audit sink. When BRAIN_LOKI_URL is
 // set it builds the classification-aware DegradingSink (loki central +
 // durable file buffer + optional ntfy) and starts the reconcile loop;
@@ -161,12 +171,7 @@ func buildAuditSink(ctx context.Context, brainDir string, logger *slog.Logger) c
 		logger.Error("capture audit buffer init", "err", err)
 		os.Exit(1)
 	}
-	// Keep notifier as a nil interface (not a typed-nil) when unconfigured
-	// so DegradingSink/Reconcile skip it cleanly.
-	var notifier audit.Notifier
-	if n := audit.NewNtfyNotifier(os.Getenv("BRAIN_NTFY_URL"), os.Getenv("BRAIN_NTFY_TOKEN")); n != nil {
-		notifier = n
-	}
+	notifier := buildNotifier()
 	reconcileInterval := time.Duration(envInt("BRAIN_AUDIT_RECONCILE_INTERVAL", 60)) * time.Second
 	audit.StartReconcile(ctx, central, buffer, notifier, reconcileInterval)
 	logger.Info("capture audit: loki+buffer sink", "loki", lokiURL, "reconcile_s", int(reconcileInterval.Seconds()))
@@ -490,7 +495,13 @@ func main() {
 		// the existing /mcp OAuth connector. mcpSrv is already wrapped above;
 		// WithCapture mutates the same instance, so the tool appears live.
 		mcpSrv.WithCapture(captureSvc, jwtValidator, mcpToken, "local-cli", resolver)
-		logger.Info("capture enabled (REST + MCP tool)", "sovereign_principals", len(sovereign))
+		// hyperguild#97: a governance-file rewrite gets the same trail a
+		// capture already gets, plus an out-of-band alert. Wired here because
+		// brain_edit_root needs the capture plumbing anyway — that is what
+		// re-derives the static principal its gate checks.
+		mcpSrv.WithGovernanceObserver(auditSink, buildNotifier())
+		logger.Info("capture enabled (REST + MCP tool)", "sovereign_principals", len(sovereign),
+			"governance_observer", true)
 	} else {
 		logger.Info("capture endpoint disabled (BRAIN_GITEA_TOKEN unset)")
 	}

ingestion/internal/api/handler.go

@@ -12,6 +12,7 @@ import (
 	"time"
 
 	"github.com/mathiasbq/hyperguild/ingestion/internal/brain"
+	"github.com/mathiasbq/hyperguild/ingestion/internal/claim"
 	"github.com/mathiasbq/hyperguild/ingestion/internal/extract"
 	"github.com/mathiasbq/hyperguild/ingestion/internal/pipeline"
 	"github.com/mathiasbq/hyperguild/ingestion/internal/search"
@@ -152,15 +153,24 @@ func writeHallNote(brainDir string, opts WriteNoteOptions) (string, error) {
 	if err != nil {
 		return "", err
 	}
-	if err := os.MkdirAll(filepath.Dir(dest), 0o755); err != nil {
-		return "", fmt.Errorf("create hall dir: %w", err)
-	}
-
 	existingFields, body := splitFrontmatter(opts.Content)
 	existingByKey := make(map[string]frontmatterField, len(existingFields))
 	for _, f := range existingFields {
 		existingByKey[f.key] = f
 	}
+
+	// Validate the claim-property envelope BEFORE creating the directory, so a
+	// rejected write leaves nothing behind — not even an empty hall dir
+	// (brain#21). Fail closed: a value outside the enum is worse than an absent
+	// one, because a lint skips what it cannot parse and the note then reads as
+	// classified while nothing acts on it.
+	if err := claim.ValidateEnvelope(scalarValues(existingFields)); err != nil {
+		return "", fmt.Errorf("claim envelope: %w", err)
+	}
+
+	if err := os.MkdirAll(filepath.Dir(dest), 0o755); err != nil {
+		return "", fmt.Errorf("create hall dir: %w", err)
+	}
 	emitted := make(map[string]bool, 6)
 
 	var fm strings.Builder
@@ -218,6 +228,25 @@ func writeHallNote(brainDir string, opts WriteNoteOptions) (string, error) {
 	return filepath.ToSlash(rel), nil
 }
 
+// scalarValues flattens frontmatter fields to key -> scalar value for
+// validation. Multi-line values (lists, nested blocks) yield the text after
+// the colon on the first line, which is empty for a list — correct here,
+// since every claim-property field is a scalar enum.
+func scalarValues(fields []frontmatterField) map[string]string {
+	out := make(map[string]string, len(fields))
+	for _, f := range fields {
+		if len(f.lines) == 0 {
+			continue
+		}
+		_, value, found := strings.Cut(f.lines[0], ":")
+		if !found {
+			continue
+		}
+		out[f.key] = strings.TrimSpace(value)
+	}
+	return out
+}
+
 // frontmatterField is one top-level YAML key from a frontmatter block,
 // along with its raw line and any indented continuation lines (e.g. a
 // bulleted list value spanning multiple lines).

ingestion/internal/api/handler_test.go

@@ -440,3 +440,60 @@ func TestIngestPath_Directory(t *testing.T) {
 	require.True(t, ok, "pages must be an array")
 	assert.NotEmpty(t, pagesSlice)
 }
+
+// --- brain#21: claim-property envelope on the write path -------------------
+
+func TestWriteNote_PreservesClaimEnvelopeVerbatim(t *testing.T) {
+	// writeHallNote already passes unknown keys through; this pins that
+	// behaviour for the envelope specifically, so a future refactor of the
+	// merge loop cannot silently drop it.
+	dir := t.TempDir()
+
+	rel, err := api.WriteNote(dir, api.WriteNoteOptions{
+		Content: "---\nhalf_life: durable\napplies: universal\nclaim_provenance: asserted\n---\n\nA durable claim.\n",
+		Wing:    "homelab",
+		Hall:    "decisions",
+	})
+	require.NoError(t, err)
+
+	got, err := os.ReadFile(filepath.Join(dir, filepath.FromSlash(rel)))
+	require.NoError(t, err)
+	assert.Contains(t, string(got), "half_life: durable")
+	assert.Contains(t, string(got), "applies: universal")
+	assert.Contains(t, string(got), "claim_provenance: asserted")
+}
+
+func TestWriteNote_RejectsInvalidClaimEnvelopeAndWritesNothing(t *testing.T) {
+	// Fail closed: a value outside the enum must not land. A lint skips what
+	// it cannot parse, so an unvalidated write produces a note that reads as
+	// classified while nothing acts on it.
+	dir := t.TempDir()
+
+	_, err := api.WriteNote(dir, api.WriteNoteOptions{
+		Content:  "---\nhalf_life: forever\nclaim_provenance: asserted\n---\n\nBody.\n",
+		Filename: "bad-envelope",
+		Wing:     "homelab",
+		Hall:     "facts",
+	})
+	require.Error(t, err)
+	assert.Contains(t, err.Error(), "half_life")
+
+	_, statErr := os.Stat(filepath.Join(dir, "wiki/homelab/facts/bad-envelope.md"))
+	assert.True(t, os.IsNotExist(statErr), "a rejected write must leave no file behind")
+}
+
+func TestWriteNote_RejectsClaimFieldWithoutProvenance(t *testing.T) {
+	dir := t.TempDir()
+
+	_, err := api.WriteNote(dir, api.WriteNoteOptions{
+		Content:  "---\napplies: homelab\n---\n\nBody.\n",
+		Filename: "no-provenance",
+		Wing:     "homelab",
+		Hall:     "facts",
+	})
+	require.Error(t, err)
+	assert.Contains(t, err.Error(), "claim_provenance")
+
+	_, statErr := os.Stat(filepath.Join(dir, "wiki/homelab/facts/no-provenance.md"))
+	assert.True(t, os.IsNotExist(statErr))
+}

ingestion/internal/capture/service.go

@@ -83,7 +83,7 @@ func (s *Service) Capture(ctx context.Context, in CaptureInput) (CaptureReceipt,
 		declared, _ = classification.ParseLevel(in.Context.Classification)
 	}
 
-	effective, securityEvents := s.resolveClassification(declared, in)
+	effective, securityEvents, escalators := s.resolveClassification(declared, in)
 
 	// Server-derived origin governs the I1 gate; a caller-asserted harness
 	// label that names a different zone is descriptive-only and logged as a
@@ -107,8 +107,8 @@ func (s *Service) Capture(ctx context.Context, in CaptureInput) (CaptureReceipt,
 			Items:                   nil, // refused before any write
 			SecurityEvents:          append(securityEvents, "I1 refusal: confidential capture via us-nexus origin"),
 		}, AuditCentral)
-		return CaptureReceipt{}, fmt.Errorf("%w: effective classification confidential through %s origin",
-			ErrSovereigntyRefused, in.Context.Origin)
+		return CaptureReceipt{}, fmt.Errorf("%w: effective classification confidential through %s origin (%s)",
+			ErrSovereigntyRefused, in.Context.Origin, escalationReason(escalators))
 	}
 
 	receipt := CaptureReceipt{
@@ -221,15 +221,28 @@ func (s *Service) validate(in CaptureInput) error {
 	return nil
 }
 
-// resolveClassification computes the effective level (stricter of
-// declared and every target's derived level) and collects a security
-// event whenever the caller under-declared relative to a target floor.
-func (s *Service) resolveClassification(declared classification.Level, in CaptureInput) (classification.Level, []string) {
+// resolveClassification computes the effective level (stricter of declared
+// and every target's derived level), collects a security event whenever the
+// caller under-declared relative to a target floor, and names the targets
+// that carry the strictest level.
+//
+// That third return is what makes an I1 refusal actionable. A caller sending
+// twenty tickets gets one refusal; without the offending target it cannot
+// tell which repo tripped the gate, and infra#265 shows what happens next —
+// the caller concluded the tickets ARRAY was broken, built a fallback of
+// twenty individual calls, and filed a bug against the wrong thing.
+func (s *Service) resolveClassification(declared classification.Level, in CaptureInput) (classification.Level, []string, []string) {
 	effective := declared
 	var events []string
+	derivedFor := map[string]classification.Level{}
+	var order []string
 	consider := func(kind classification.TargetKind, name string) {
 		derived := s.policy.Derive(classification.Target{Kind: kind, Name: name})
 		effective = classification.Stricter(effective, derived)
+		if _, seen := derivedFor[name]; !seen {
+			derivedFor[name] = derived
+			order = append(order, name)
+		}
 		if declared < derived {
 			events = append(events, fmt.Sprintf("classification under-declared: declared=%s target=%s(%s) derived=%s",
 				declared, name, kindString(kind), derived))
@@ -241,7 +254,28 @@ func (s *Service) resolveClassification(declared classification.Level, in Captur
 	for _, tk := range in.Tickets {
 		consider(classification.RepoTarget, tk.Repo)
 	}
-	return effective, events
+
+	// Only targets that actually reach the effective level are named: at
+	// twenty tickets, listing the innocent ones is noise that hides the one
+	// that matters.
+	var escalators []string
+	for _, name := range order {
+		if derivedFor[name] == effective && effective > declared {
+			escalators = append(escalators, name)
+		}
+	}
+	return effective, events, escalators
+}
+
+// escalationReason renders why the effective classification landed where it
+// did, for an operator reading a refusal. Naming the caller's own declaration
+// matters as much as naming a target: "you asked for this" and "a repo you
+// touched required it" call for different fixes.
+func escalationReason(escalators []string) string {
+	if len(escalators) == 0 {
+		return "declared by the caller; no target required it"
+	}
+	return "required by: " + strings.Join(escalators, ", ")
 }
 
 func (s *Service) persistInsight(ctx context.Context, ins Insight) (InsightResult, string, error) {

ingestion/internal/capture/service_test.go

@@ -3,6 +3,7 @@ package capture
 import (
 	"context"
 	"errors"
+	"fmt"
 	"strings"
 	"testing"
 	"time"
@@ -59,12 +60,18 @@ type fakeTracker struct {
 	closed   []int
 	comments []int
 	err      error
+	// failOnTitle makes CreateIssue fail for one specific title, so a
+	// partial-failure receipt can be exercised without failing every item.
+	failOnTitle string
 }
 
 func (f *fakeTracker) CreateIssue(_ context.Context, repo, title, _ string) (IssueRef, error) {
 	if f.err != nil {
 		return IssueRef{}, f.err
 	}
+	if f.failOnTitle != "" && title == f.failOnTitle {
+		return IssueRef{}, errors.New("tracker refused this one")
+	}
 	f.created = append(f.created, repo+":"+title)
 	return IssueRef{Repo: repo, Number: 100 + len(f.created), URL: "https://git/" + repo + "/issues/x"}, nil
 }
@@ -87,7 +94,9 @@ func (f *fakeTracker) CommentIssue(_ context.Context, repo string, number int, _
 
 // fakePolicy derives from an explicit map; default Internal so tests pin
 // behaviour without depending on the real defaulting.
-type fakePolicy struct{ tags map[string]classification.Level }
+type fakePolicy struct {
+	tags map[string]classification.Level
+}
 
 func (p fakePolicy) Derive(t classification.Target) classification.Level {
 	if lvl, ok := p.tags[t.Name]; ok {
@@ -434,3 +443,98 @@ func TestCaptureDryRunSkipsAuditGate(t *testing.T) {
 	assert.True(t, rec.DryRun)
 	assert.Empty(t, au.entries)
 }
+
+// ---- hyperguild#91 part 1 -------------------------------------------------
+//
+// The reported symptom was "capture with a tickets array fails" (infra#265:
+// 20 tickets errored, 5 errored, caller fell back to 20 individual
+// issue_create calls). The array was never the failing dimension — every
+// ticket targeted a client repo, and the I1 gate refuses that through a
+// us-nexus origin at ANY length. The refusal did not say which target
+// escalated the classification, so it read as a malfunction.
+
+func TestCapture_TicketsArrayWorksAtLengthOne(t *testing.T) {
+	b, tr, au := &fakeBrain{}, &fakeTracker{}, &fakeAudit{}
+	svc := newSvc(b, tr, fakePolicy{}, au)
+
+	rec, err := svc.Capture(context.Background(), CaptureInput{
+		Context: baseCtx(),
+		Tickets: []Ticket{{Repo: "hyperguild", Action: "create", Title: "one", Body: "b"}},
+	})
+	require.NoError(t, err)
+	require.Len(t, rec.Tickets, 1)
+	assert.True(t, rec.Tickets[0].OK)
+	assert.Len(t, tr.created, 1)
+	assert.Empty(t, rec.Errors)
+}
+
+func TestCapture_TicketsArrayWorksAtLengthMany(t *testing.T) {
+	b, tr, au := &fakeBrain{}, &fakeTracker{}, &fakeAudit{}
+	svc := newSvc(b, tr, fakePolicy{}, au)
+
+	in := CaptureInput{Context: baseCtx()}
+	for i := range 20 {
+		in.Tickets = append(in.Tickets, Ticket{
+			Repo: "hyperguild", Action: "create", Title: fmt.Sprintf("t%02d", i), Body: "b",
+		})
+	}
+
+	rec, err := svc.Capture(context.Background(), in)
+	require.NoError(t, err)
+	require.Len(t, rec.Tickets, 20, "one receipt entry per ticket")
+	for i, r := range rec.Tickets {
+		assert.True(t, r.OK, "ticket %d failed: %+v", i, rec.Errors)
+	}
+	assert.Len(t, tr.created, 20)
+	assert.Empty(t, rec.Errors)
+}
+
+func TestCapture_OneFailingTicketDoesNotSinkTheRest(t *testing.T) {
+	// infra#265 asked explicitly: "if ticket 3 of 5 fails, what happened to
+	// 1 and 2?" They land, and the receipt says so per item.
+	b, au := &fakeBrain{}, &fakeAudit{}
+	tr := &fakeTracker{failOnTitle: "boom"}
+	svc := newSvc(b, tr, fakePolicy{}, au)
+
+	rec, err := svc.Capture(context.Background(), CaptureInput{
+		Context: baseCtx(),
+		Tickets: []Ticket{
+			{Repo: "hyperguild", Action: "create", Title: "ok-1", Body: "b"},
+			{Repo: "hyperguild", Action: "create", Title: "boom", Body: "b"},
+			{Repo: "hyperguild", Action: "create", Title: "ok-3", Body: "b"},
+		},
+	})
+	require.NoError(t, err, "a per-item failure is a partial receipt, not a call error")
+	require.Len(t, rec.Tickets, 3)
+	assert.True(t, rec.Tickets[0].OK)
+	assert.False(t, rec.Tickets[1].OK)
+	assert.True(t, rec.Tickets[2].OK)
+	require.Len(t, rec.Errors, 1)
+	assert.Contains(t, rec.Errors[0].Item, "ticket[1]")
+}
+
+func TestCapture_I1RefusalNamesWhatEscalatedTheClassification(t *testing.T) {
+	// The actual defect behind infra#265. Without the offending target in
+	// the message, a caller with 20 tickets cannot tell which repo tripped
+	// the gate — and reasonably concludes the array itself is broken.
+	b, tr, au := &fakeBrain{}, &fakeTracker{}, &fakeAudit{}
+	pol := fakePolicy{tags: map[string]classification.Level{"trucker-ecommerce": classification.Confidential}}
+	svc := newSvc(b, tr, pol, au)
+
+	ctx := baseCtx()
+	ctx.Origin = ZoneUSNexus
+	_, err := svc.Capture(context.Background(), CaptureInput{
+		Context: ctx,
+		Tickets: []Ticket{
+			{Repo: "hyperguild", Action: "create", Title: "fine", Body: "b"},
+			{Repo: "trucker-ecommerce", Action: "create", Title: "client work", Body: "b"},
+		},
+	})
+
+	require.ErrorIs(t, err, ErrSovereigntyRefused)
+	assert.Contains(t, err.Error(), "trucker-ecommerce",
+		"the refusal must name the target that escalated the classification")
+	assert.NotContains(t, err.Error(), "hyperguild",
+		"only the offending target, or the message is noise at 20 tickets")
+	assert.Empty(t, tr.created, "refused before any write")
+}

ingestion/internal/claim/claim.go

@@ -0,0 +1,98 @@
+// Package claim validates the claim-property frontmatter that says when and
+// where a note's claim holds, and who decided that (brain#21).
+//
+// Two families were designed together and land separately. This is the
+// applicability envelope:
+//
+//	half_life:        volatile | seasonal | durable   — how fast it decays
+//	applies:          local | homelab | universal     — how widely it holds
+//	claim_provenance: asserted | inferred             — who set the above
+//
+// The strength family (confidence, contested) follows once the ingest-time
+// contradiction detector exists to populate it.
+//
+// Why the envelope is validated rather than merely accepted: a value outside
+// the enum is worse than an absent one. A lint skips what it cannot parse, so
+// the note reads as classified while nothing acts on it — the field becomes
+// decoration that suppresses the very review it was added to trigger.
+//
+// Why `applies` and not `scope`: `scope:` has been in use since brain#5 for
+// glossary terms (`homelab-core`, `project:<repo>`) — a disjoint value space
+// on the same key. The shipped schema keeps the name; this one takes another.
+package claim
+
+import (
+	"fmt"
+	"strings"
+)
+
+// Field names, exported so writers and lints agree on the spelling.
+const (
+	FieldHalfLife   = "half_life"
+	FieldApplies    = "applies"
+	FieldProvenance = "claim_provenance"
+)
+
+var (
+	halfLives   = []string{"volatile", "seasonal", "durable"}
+	appliesTo   = []string{"local", "homelab", "universal"}
+	provenances = []string{"asserted", "inferred"}
+)
+
+// ValidateEnvelope checks the envelope fields of one note's frontmatter.
+//
+// Absent fields are valid: the envelope is opt-in for new writes, and a
+// corpus-wide backfill is deliberately out of scope (brain#21) — inferring
+// 1113 notes' properties in bulk has the same failure mode as the extraction
+// pass that produced 28 wrong wikilinks out of 67.
+func ValidateEnvelope(fm map[string]string) error {
+	halfLife, hasHalfLife := present(fm, FieldHalfLife)
+	applies, hasApplies := present(fm, FieldApplies)
+	prov, hasProv := present(fm, FieldProvenance)
+
+	if hasHalfLife {
+		if err := oneOf(FieldHalfLife, halfLife, halfLives); err != nil {
+			return err
+		}
+	}
+	if hasApplies {
+		if err := oneOf(FieldApplies, applies, appliesTo); err != nil {
+			return err
+		}
+	}
+	if hasProv {
+		if err := oneOf(FieldProvenance, prov, provenances); err != nil {
+			return err
+		}
+	}
+
+	// The field and its provenance travel together or not at all. A value
+	// whose author is unknown cannot be trusted differently from one Mathias
+	// asserted, which is how an inferred guess hardens into accepted fact.
+	switch {
+	case (hasHalfLife || hasApplies) && !hasProv:
+		return fmt.Errorf("%s is required when %s or %s is set",
+			FieldProvenance, FieldHalfLife, FieldApplies)
+	case hasProv && !hasHalfLife && !hasApplies:
+		return fmt.Errorf("%s is set but neither %s nor %s is",
+			FieldProvenance, FieldHalfLife, FieldApplies)
+	}
+	return nil
+}
+
+// present reports a trimmed field value and whether it was set to anything
+// non-empty. A key written with an empty value is treated as absent rather
+// than as an error: that is a caller emitting a placeholder, not asserting.
+func present(fm map[string]string, key string) (string, bool) {
+	v := strings.TrimSpace(fm[key])
+	return v, v != ""
+}
+
+func oneOf(field, value string, allowed []string) error {
+	for _, a := range allowed {
+		if a == value {
+			return nil
+		}
+	}
+	return fmt.Errorf("%s: %q is not one of %s", field, value, strings.Join(allowed, ", "))
+}

ingestion/internal/claim/claim_test.go

@@ -0,0 +1,89 @@
+// Package claim_test covers the claim-property envelope (brain#21).
+package claim_test
+
+import (
+	"testing"
+
+	"github.com/mathiasbq/hyperguild/ingestion/internal/claim"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestValidateEnvelope(t *testing.T) {
+	tests := []struct {
+		name    string
+		fm      map[string]string
+		wantErr string
+	}{
+		{
+			name: "absent is fine — the fields are opt-in on new writes",
+			fm:   map[string]string{"wing": "homelab", "hall": "facts"},
+		},
+		{
+			name: "complete envelope with provenance",
+			fm: map[string]string{
+				"half_life": "durable", "applies": "universal", "claim_provenance": "asserted",
+			},
+		},
+		{
+			name: "inferred provenance is equally valid — it just must be stated",
+			fm: map[string]string{
+				"half_life": "volatile", "applies": "local", "claim_provenance": "inferred",
+			},
+		},
+		{
+			// A value nobody can act on is worse than no value: a lint
+			// silently skips it, so the note looks classified and isn't.
+			name:    "unknown half_life value",
+			fm:      map[string]string{"half_life": "forever", "claim_provenance": "asserted"},
+			wantErr: `half_life: "forever" is not one of volatile, seasonal, durable`,
+		},
+		{
+			name:    "unknown applies value",
+			fm:      map[string]string{"applies": "galaxy", "claim_provenance": "asserted"},
+			wantErr: `applies: "galaxy" is not one of local, homelab, universal`,
+		},
+		{
+			// brain#21: a value a model inferred must never be
+			// indistinguishable from one Mathias asserted. Writing the field
+			// without saying who set it is exactly that laundering.
+			name:    "envelope field without provenance",
+			fm:      map[string]string{"half_life": "durable"},
+			wantErr: "claim_provenance is required when half_life or applies is set",
+		},
+		{
+			name:    "applies without provenance",
+			fm:      map[string]string{"applies": "homelab"},
+			wantErr: "claim_provenance is required when half_life or applies is set",
+		},
+		{
+			name:    "unknown provenance value",
+			fm:      map[string]string{"half_life": "durable", "claim_provenance": "probably"},
+			wantErr: `claim_provenance: "probably" is not one of asserted, inferred`,
+		},
+		{
+			// Provenance alone says who set fields that aren't there.
+			name:    "provenance without any envelope field",
+			fm:      map[string]string{"claim_provenance": "asserted"},
+			wantErr: "claim_provenance is set but neither half_life nor applies is",
+		},
+		{
+			// The glossary wing has used scope: homelab-core since #5. The
+			// envelope deliberately does NOT claim that key (brain#21).
+			name: "glossary scope is untouched",
+			fm:   map[string]string{"scope": "homelab-core", "term": "koala"},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			err := claim.ValidateEnvelope(tt.fm)
+			if tt.wantErr == "" {
+				assert.NoError(t, err)
+				return
+			}
+			require.Error(t, err)
+			assert.Contains(t, err.Error(), tt.wantErr)
+		})
+	}
+}

ingestion/internal/mcp/brain_update_test.go

@@ -165,7 +165,7 @@ func TestBrainUpdateReembedsNewContent(t *testing.T) {
 	ctx := context.Background()
 
 	callResult(t, toolCall(t, srv, "brain_write", map[string]any{
-		"content": "# Note\n\nthe OLD distinctive payload\n",
+		"content":  "# Note\n\nthe OLD distinctive payload\n",
 		"filename": "n", "wing": "a", "hall": "facts",
 	}))
 

ingestion/internal/mcp/handlers.go

@@ -14,6 +14,7 @@ import (
 	"github.com/mathiasbq/hyperguild/ingestion/internal/capture"
 	"github.com/mathiasbq/hyperguild/ingestion/internal/extract"
 	"github.com/mathiasbq/hyperguild/ingestion/internal/graphsync"
+	"github.com/mathiasbq/hyperguild/ingestion/internal/notes"
 	"github.com/mathiasbq/hyperguild/ingestion/internal/pipeline"
 	"github.com/mathiasbq/hyperguild/ingestion/internal/search"
 	"github.com/mathiasbq/hyperguild/ingestion/internal/session"
@@ -97,6 +98,8 @@ func (s *Server) tools() []map[string]any {
 				"slug":     str("optional target slug; defaults to filename minus date prefix"),
 			}),
 		},
+		noteToolDescriptors()[0],
+		noteToolDescriptors()[1],
 		{
 			"name":        "brain_tunnel",
 			"description": "Create an explicit bidirectional [[wikilink]] between two notes in different wings. Idempotent.",
@@ -320,6 +323,15 @@ func (s *Server) brainGet(ctx context.Context, args json.RawMessage) (json.RawMe
 	if target == "" {
 		return nil, fmt.Errorf("id or path is required")
 	}
+	// Containment: the store resolves any brain-relative path, and
+	// parseFrontmatter returns the WHOLE file as body when there is no
+	// frontmatter fence — so without this, brain_get reads verbatim session
+	// transcripts under archive/ and sessions/, which BM25 never indexes and
+	// which nobody therefore thinks of as reachable. Same allowlist as
+	// brain_read (brain#4).
+	if _, err := notes.Resolve(s.brainDir, target, notes.ScopeNotes); err != nil {
+		return nil, err
+	}
 	note, err := s.store.Get(ctx, target)
 	if err != nil {
 		return nil, err

ingestion/internal/mcp/server.go

@@ -49,6 +49,8 @@ type rpcError struct {
 // Server handles MCP JSON-RPC over HTTP for the ingestion service.
 type Server struct {
 	brainDir  string
+	govAudit  capture.AuditSink
+	govNotify govNotifier
 	pipeline  pipeline.Config
 	llm       pipeline.CompleteFunc
 	answerLLM pipeline.CompleteFunc // nil = brain_answer and brain_classify unavailable
@@ -275,6 +277,15 @@ func (s *Server) handleCall(ctx context.Context, name string, args json.RawMessa
 		return s.brainPromote(ctx, args)
 	case "capture":
 		return s.brainCapture(ctx, args)
+	case "brain_read":
+		return s.brainRead(ctx, args)
+	case "brain_edit":
+		return s.brainEdit(ctx, args)
+	case "brain_edit_root":
+		// Dispatches but is intentionally absent from tools/list — see
+		// tools_notes.go for why a governance-file editor must not be
+		// discoverable by a model.
+		return s.brainEditRoot(ctx, args)
 	case "brain_index":
 		return s.brainIndex(ctx, args)
 	case "brain_tunnel":

ingestion/internal/mcp/server_test.go

@@ -59,6 +59,7 @@ func TestServerToolsList(t *testing.T) {
 	assert.ElementsMatch(t, []string{
 		"brain_query", "brain_write", "brain_update", "brain_get",
 		"brain_pending", "brain_promote",
+		"brain_read", "brain_edit",
 		"brain_tunnel",
 		"brain_ingest_raw", "brain_ingest",
 		"brain_answer", "brain_classify", "brain_graph", "brain_context",
@@ -69,6 +70,13 @@ func TestServerToolsList(t *testing.T) {
 	// a caller verb is cognitive load with no decision behind it (ADR-0012
 	// rule 4, #78). The capability stays reachable — see the admin test below.
 	assert.NotContains(t, names, "brain_index")
+
+	// brain_edit_root is de-surfaced for a sharper reason than brain_index:
+	// brain content is ingested from the outside world and reaches a model's
+	// context, so a listed governance-file editor would be a durable
+	// prompt-injection path — one poisoned page rewriting the CLAUDE.md a
+	// .dispatch-allow agent later loads as instructions.
+	assert.NotContains(t, names, "brain_edit_root")
 }
 
 func TestBrainIndexStaysCallableForAdminUseAfterDeSurfacing(t *testing.T) {

ingestion/internal/mcp/tools_capture.go

@@ -31,7 +31,7 @@ func captureToolDescriptor() map[string]any {
 		"type": "object",
 		"properties": map[string]any{
 			"text": str("the insight body"), "wing": str("brain wing"),
-			"hall": str("brain hall (facts/decisions/failures/hypotheses/sources)"),
+			"hall":           str("brain hall (facts/decisions/failures/hypotheses/sources)"),
 			"supersede_slug": str("optional: slug of a prior note to revise in place instead of creating"),
 		},
 		"required": []string{"text", "wing", "hall"},
@@ -41,7 +41,7 @@ func captureToolDescriptor() map[string]any {
 		"properties": map[string]any{
 			"repo": str("gitea repo (owner is always mathias)"), "action": str("create|close|comment"),
 			"number": map[string]any{"type": "integer", "description": "issue number (close/comment)"},
-			"title": str("issue title (create)"), "body": str("issue/comment body"),
+			"title":  str("issue title (create)"), "body": str("issue/comment body"),
 		},
 		"required": []string{"repo", "action"},
 	}
@@ -51,14 +51,14 @@ func captureToolDescriptor() map[string]any {
 			"context": map[string]any{
 				"type": "object",
 				"properties": map[string]any{
-					"harness": str("descriptive harness label (telemetry only, never a gate input)"),
+					"harness":     str("descriptive harness label (telemetry only, never a gate input)"),
 					"session_ref": str("optional session reference"), "fidelity": str("live-capture|transcript-parse|agent-runlog"),
 					"actor": str("acting user/agent"), "classification": str("caller-declared sensitivity: public|internal|confidential"),
 				},
 			},
 			"insights": map[string]any{"type": "array", "items": insightItem},
 			"tickets":  map[string]any{"type": "array", "items": ticketItem},
-			"dry_run": map[string]any{"type": "boolean", "description": "validate + return the would-be receipt, write nothing"},
+			"dry_run":  map[string]any{"type": "boolean", "description": "validate + return the would-be receipt, write nothing"},
 		},
 	}
 	b, _ := json.Marshal(schema)

ingestion/internal/mcp/tools_notes.go

@@ -0,0 +1,222 @@
+package mcp
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"log/slog"
+	"strings"
+	"time"
+
+	"github.com/mathiasbq/hyperguild/ingestion/internal/brain"
+	"github.com/mathiasbq/hyperguild/ingestion/internal/capture"
+	"github.com/mathiasbq/hyperguild/ingestion/internal/classification"
+	"github.com/mathiasbq/hyperguild/ingestion/internal/notes"
+)
+
+// noteToolDescriptors are the model-facing halves of brain#4.
+//
+// brain_edit_root is deliberately absent: it dispatches but is not listed, the
+// same de-surfacing ADR-0012 rule 4 applies to brain_index (#78). Hiding the
+// verb matters here for a specific reason — brain content is ingested from the
+// outside world and reaches a model's context, so a listed governance-file
+// editor is a durable prompt-injection path: one poisoned page could rewrite
+// the CLAUDE.md that a .dispatch-allow agent later obeys. A human who knows
+// the tool exists can still call it; a model reading tools/list cannot
+// discover it.
+func noteToolDescriptors() []map[string]any {
+	str := func(d string) map[string]any { return map[string]any{"type": "string", "description": d} }
+	schema := func(required []string, props map[string]any) json.RawMessage {
+		b, _ := json.Marshal(map[string]any{"type": "object", "required": required, "properties": props})
+		return b
+	}
+	return []map[string]any{
+		{
+			"name": "brain_read",
+			"description": "Return a note's raw markdown verbatim — no synthesis, no excerpting. " +
+				"Use this to obtain the exact text before brain_edit; never build old_str from a " +
+				"brain_query excerpt, which is truncated and normalised and will not match.",
+			"inputSchema": schema([]string{"path"}, map[string]any{
+				"path": str("brain-relative path under knowledge/ or wiki/, e.g. wiki/homelab/facts/x.md"),
+			}),
+		},
+		{
+			"name": "brain_edit",
+			"description": "Replace exactly one occurrence of old_str with new_str in a note, leaving " +
+				"every other byte identical. Fails loudly if old_str is absent or matches more than " +
+				"once — extend old_str until it is unique. Does NOT stamp updated_at or supersedes " +
+				"(brain_update does); it is a byte-level patch, not a metadata operation.",
+			"inputSchema": schema([]string{"path", "old_str", "new_str"}, map[string]any{
+				"path":    str("brain-relative path under knowledge/ or wiki/"),
+				"old_str": str("exact text to replace; must occur exactly once"),
+				"new_str": str("replacement text; empty string deletes old_str"),
+			}),
+		},
+	}
+}
+
+// --- hyperguild#97: governance-rewrite observability -----------------------
+
+// govNotifier raises an out-of-band alert. Declared here rather than importing
+// internal/audit so the MCP package keeps depending on capture alone.
+type govNotifier interface {
+	Notify(ctx context.Context, msg string) error
+}
+
+// WithGovernanceObserver attaches the audit sink and notifier that observe a
+// successful brain_edit_root. Both are optional; either may be nil.
+func (s *Server) WithGovernanceObserver(sink capture.AuditSink, n govNotifier) {
+	s.govAudit, s.govNotify = sink, n
+}
+
+type brainReadArgs struct {
+	Path string `json:"path"`
+}
+
+func (s *Server) brainRead(_ context.Context, args json.RawMessage) (json.RawMessage, error) {
+	var a brainReadArgs
+	if err := json.Unmarshal(args, &a); err != nil {
+		return nil, fmt.Errorf("parse args: %w", err)
+	}
+	content, err := notes.Read(s.brainDir, a.Path)
+	if err != nil {
+		return nil, err
+	}
+	return json.Marshal(map[string]any{"path": a.Path, "content": string(content)})
+}
+
+type brainEditArgs struct {
+	Path   string `json:"path"`
+	OldStr string `json:"old_str"`
+	NewStr string `json:"new_str"`
+}
+
+// brainEdit is the model-facing editor: notes only.
+//
+// Deliberately NOT audited (#97). A note edit is already fully reconstructible
+// from git history, which carries the same before/after bytes plus a diff.
+// Governance files get the extra trail because they change what later agents
+// obey — the asymmetry is about instruction-carrying content, not about how
+// hard the change is to reconstruct. Auditing every note edit would also bury
+// the rare governance record under routine traffic, which is how an alert
+// stops being read.
+func (s *Server) brainEdit(_ context.Context, args json.RawMessage) (json.RawMessage, error) {
+	_, out, err := s.editWithScope(args, notes.ScopeNotes)
+	return out, err
+}
+
+// brainEditRoot edits a governance file (CLAUDE.md). De-surfaced AND gated on
+// the authenticated principal: only the static local-CLI token may call it, so
+// a claude.ai-origin session — the one that may have just read an untrusted
+// page — cannot rewrite the instructions later agents load.
+func (s *Server) brainEditRoot(ctx context.Context, args json.RawMessage) (json.RawMessage, error) {
+	info, ok := ctx.Value(principalKey).(principalInfo)
+	if !ok || !info.viaStatic {
+		return nil, fmt.Errorf("brain_edit_root requires the local static principal; " +
+			"a governance file is not editable from a model-facing origin")
+	}
+	res, out, err := s.editWithScope(args, notes.ScopeRoot)
+	if err != nil {
+		// A refusal is not a rewrite. Recording or alerting on one would cry
+		// wolf, and the gate already logs its own denials.
+		return out, err
+	}
+	// A durable instruction rewrite should be loud, not silent.
+	slog.Warn("brain_edit_root applied", "principal", info.principal, "result", string(out))
+	s.observeGovernanceEdit(ctx, info.principal, res)
+	return out, err
+}
+
+// observeGovernanceEdit records and alerts on a rewrite that already landed.
+//
+// Best-effort by construction: the bytes are written and brain-sync will commit
+// them within the quarter hour, so failing the call here would report a
+// non-rewrite to the caller while the rewrite sits on disk. A failure to
+// observe is therefore logged at ERROR — the one case where the trail itself is
+// what broke, and the only remaining signal.
+func (s *Server) observeGovernanceEdit(ctx context.Context, principal string, res notes.Result) {
+	// Path and hashes only. The notification leaves the cluster for a phone
+	// via a third-party topic; a hash identifies the rewrite, the bytes would
+	// leak whatever was written into an instruction file.
+	msg := fmt.Sprintf("brain_edit_root applied\npath=%s\nprincipal=%s\nbefore=%s\nafter=%s",
+		res.Path, principal, res.PrevContentHash, res.ContentHash)
+
+	if s.govNotify != nil {
+		if err := s.govNotify.Notify(ctx, msg); err != nil {
+			slog.Error("brain_edit_root notify failed", "path", res.Path, "err", err)
+		}
+	}
+	if s.govAudit == nil {
+		return
+	}
+	// Internal, not Confidential: a governance file is not client data, and
+	// Confidential would make DegradingSink REFUSE to record while loki is
+	// down rather than buffer — losing exactly the record that matters most.
+	outcome, err := s.govAudit.Reserve(ctx, classification.Internal)
+	if err != nil {
+		slog.Error("brain_edit_root audit reserve failed", "path", res.Path, "err", err)
+		return
+	}
+	entry := capture.AuditEntry{
+		Timestamp:               time.Now().UTC(),
+		Principal:               principal,
+		Actor:                   "brain_edit_root",
+		Harness:                 "mcp",
+		EffectiveClassification: classification.Internal.String(),
+		Items:                   []string{res.Path},
+		SecurityEvents: []string{fmt.Sprintf(
+			"governance file rewritten: %s before=%s after=%s",
+			res.Path, res.PrevContentHash, res.ContentHash)},
+	}
+	if err := s.govAudit.Record(ctx, entry, outcome); err != nil {
+		slog.Error("brain_edit_root audit record failed", "path", res.Path, "err", err)
+	}
+}
+
+func (s *Server) editWithScope(args json.RawMessage, scope notes.Scope) (notes.Result, json.RawMessage, error) {
+	var a brainEditArgs
+	if err := json.Unmarshal(args, &a); err != nil {
+		return notes.Result{}, nil, fmt.Errorf("parse args: %w", err)
+	}
+	res, err := notes.Edit(s.brainDir, a.Path, a.OldStr, a.NewStr, scope)
+	if err != nil {
+		return notes.Result{}, nil, err
+	}
+
+	// Reindex only when the edit could have changed what the index reads.
+	// BuildWingIndex consumes title/created_at from the frontmatter block;
+	// BM25 walks the filesystem live and embeddings are mtime-driven, so an
+	// unconditional rebuild would buy nothing and dirty a second file —
+	// _index.md being precisely the class brain-sync already needs a bespoke
+	// conflict resolver for.
+	if wing := wingOf(res.Path); wing != "" && touchesFrontmatter(a.OldStr, a.NewStr) {
+		if ierr := brain.BuildWingIndex(s.brainDir, wing); ierr != nil {
+			// Best-effort, matching every other index hook here: the write
+			// already landed and will be committed, so failing the call would
+			// make the caller retry an edit whose old_str no longer exists.
+			res.ReindexError = ierr.Error()
+		}
+	}
+	out, merr := json.Marshal(res)
+	return res, out, merr
+}
+
+// wingOf returns the wing for a wiki path, or "" for knowledge/ and root.
+func wingOf(rel string) string {
+	parts := strings.Split(rel, "/")
+	if len(parts) < 3 || parts[0] != "wiki" {
+		return ""
+	}
+	return parts[1]
+}
+
+// touchesFrontmatter reports whether either side of the edit mentions a field
+// the wing index reads.
+func touchesFrontmatter(oldStr, newStr string) bool {
+	for _, s := range []string{oldStr, newStr} {
+		if strings.Contains(s, "title:") || strings.Contains(s, "created_at:") || strings.Contains(s, "---") {
+			return true
+		}
+	}
+	return false
+}

ingestion/internal/mcp/tools_notes_test.go

@@ -0,0 +1,297 @@
+// ingestion/internal/mcp/tools_notes_test.go
+package mcp_test
+
+import (
+	"context"
+	"crypto/sha256"
+	"encoding/hex"
+	"encoding/json"
+	"net/http"
+	"net/http/httptest"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"github.com/mathiasbq/hyperguild/ingestion/internal/audit"
+	"github.com/mathiasbq/hyperguild/ingestion/internal/brainstore"
+	"github.com/mathiasbq/hyperguild/ingestion/internal/capture"
+	"github.com/mathiasbq/hyperguild/ingestion/internal/capturehttp"
+	"github.com/mathiasbq/hyperguild/ingestion/internal/classification"
+	"github.com/mathiasbq/hyperguild/ingestion/internal/mcp"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// ---- brain#4: brain_read / brain_edit through the JSON-RPC surface --------
+
+func newBrainRepo(t *testing.T) string {
+	t.Helper()
+	dir := t.TempDir()
+	require.NoError(t, os.MkdirAll(filepath.Join(dir, ".git"), 0o755))
+	require.NoError(t, os.MkdirAll(filepath.Join(dir, "wiki/homelab/facts"), 0o755))
+	require.NoError(t, os.WriteFile(filepath.Join(dir, "wiki/homelab/facts/n.md"),
+		[]byte("---\ntitle: 'n'\n---\n\nalpha\nbeta\n"), 0o644))
+	require.NoError(t, os.WriteFile(filepath.Join(dir, "CLAUDE.md"), []byte("# rules\n"), 0o644))
+	return dir
+}
+
+func call(t *testing.T, srv *mcp.Server, name string, args map[string]any) map[string]any {
+	t.Helper()
+	req := httptest.NewRequest(http.MethodPost, "/mcp", body(t, map[string]any{
+		"jsonrpc": "2.0", "id": 1, "method": "tools/call",
+		"params": map[string]any{"name": name, "arguments": args},
+	}))
+	rr := httptest.NewRecorder()
+	srv.ServeHTTP(rr, req)
+	var resp map[string]any
+	require.NoError(t, json.Unmarshal(rr.Body.Bytes(), &resp))
+	return resp
+}
+
+func TestBrainReadAndEdit_RoundTripThroughJSONRPC(t *testing.T) {
+	// The Go functions are tested in internal/notes; this pins the transport,
+	// because a tool that works in-process and is unreachable over JSON-RPC
+	// is indistinguishable from a working one in a unit test.
+	dir := newBrainRepo(t)
+	srv := mcp.NewServer(dir, nil, nil, nil)
+
+	resp := call(t, srv, "brain_read", map[string]any{"path": "wiki/homelab/facts/n.md"})
+	require.NotContains(t, resp, "error", "%v", resp["error"])
+
+	resp = call(t, srv, "brain_edit", map[string]any{
+		"path": "wiki/homelab/facts/n.md", "old_str": "beta", "new_str": "BETA",
+	})
+	require.NotContains(t, resp, "error", "%v", resp["error"])
+
+	got, err := os.ReadFile(filepath.Join(dir, "wiki/homelab/facts/n.md"))
+	require.NoError(t, err)
+	assert.Equal(t, "---\ntitle: 'n'\n---\n\nalpha\nBETA\n", string(got),
+		"only the matched bytes may change")
+}
+
+func TestBrainEdit_RefusesGovernanceFilesFromTheModelFacingTool(t *testing.T) {
+	dir := newBrainRepo(t)
+	srv := mcp.NewServer(dir, nil, nil, nil)
+
+	resp := call(t, srv, "brain_edit", map[string]any{
+		"path": "CLAUDE.md", "old_str": "# rules", "new_str": "# rewritten",
+	})
+	require.Contains(t, resp, "error", "brain_edit must not reach CLAUDE.md")
+
+	got, err := os.ReadFile(filepath.Join(dir, "CLAUDE.md"))
+	require.NoError(t, err)
+	assert.Equal(t, "# rules\n", string(got))
+}
+
+func TestBrainEditRoot_RefusedWithoutTheLocalStaticPrincipal(t *testing.T) {
+	// No principal in context is the model-facing case (and the default in
+	// this test server): the privileged door must stay shut.
+	dir := newBrainRepo(t)
+	srv := mcp.NewServer(dir, nil, nil, nil)
+
+	resp := call(t, srv, "brain_edit_root", map[string]any{
+		"path": "CLAUDE.md", "old_str": "# rules", "new_str": "# rewritten",
+	})
+	require.Contains(t, resp, "error")
+
+	got, err := os.ReadFile(filepath.Join(dir, "CLAUDE.md"))
+	require.NoError(t, err)
+	assert.Equal(t, "# rules\n", string(got), "governance file must be untouched")
+}
+
+func TestBrainGet_CannotReachSessionTranscripts(t *testing.T) {
+	// Pre-existing hole, not introduced here: api.ReadNote resolves any
+	// brain-relative path and parseFrontmatter returns the WHOLE file as body
+	// when there is no frontmatter fence. archive/ holds verbatim Claude
+	// session transcripts (up to 500 KB) that BM25 never indexes, so they are
+	// invisible to search and trivially readable by path — and CLAUDE.md
+	// documents the layout.
+	dir := newBrainRepo(t)
+	require.NoError(t, os.MkdirAll(filepath.Join(dir, "archive/claude-sessions/koala"), 0o755))
+	require.NoError(t, os.WriteFile(
+		filepath.Join(dir, "archive/claude-sessions/koala/session-1.md"),
+		[]byte("verbatim transcript with tool output\n"), 0o644))
+
+	srv := mcp.NewServer(dir, nil, nil, nil)
+	resp := call(t, srv, "brain_get", map[string]any{"path": "archive/claude-sessions/koala/session-1.md"})
+
+	require.Contains(t, resp, "error", "brain_get must not read outside knowledge/ and wiki/")
+	assert.NotContains(t, string(mustJSON(t, resp)), "verbatim transcript")
+}
+
+func TestBrainGet_StillReadsNotes(t *testing.T) {
+	dir := newBrainRepo(t)
+	srv := mcp.NewServer(dir, nil, nil, nil)
+
+	resp := call(t, srv, "brain_get", map[string]any{"path": "wiki/homelab/facts/n.md"})
+	assert.NotContains(t, resp, "error", "%v", resp["error"])
+}
+
+func mustJSON(t *testing.T, v any) []byte {
+	t.Helper()
+	b, err := json.Marshal(v)
+	require.NoError(t, err)
+	return b
+}
+
+// ---- hyperguild#97: a governance rewrite must be observed, not just gated ----
+
+// govSink captures audit records. Governance edits are rare, so the whole
+// history is kept and asserted on rather than sampled.
+type govSink struct {
+	entries  []capture.AuditEntry
+	outcomes []capture.AuditOutcome
+	recErr   error
+}
+
+func (s *govSink) Reserve(context.Context, classification.Level) (capture.AuditOutcome, error) {
+	return capture.AuditCentral, nil
+}
+
+func (s *govSink) Record(_ context.Context, e capture.AuditEntry, o capture.AuditOutcome) error {
+	s.entries = append(s.entries, e)
+	s.outcomes = append(s.outcomes, o)
+	return s.recErr
+}
+
+type govNotifier struct {
+	msgs []string
+	err  error
+}
+
+func (n *govNotifier) Notify(_ context.Context, msg string) error {
+	n.msgs = append(n.msgs, msg)
+	return n.err
+}
+
+// govServer wires capture (which is what re-derives the static principal) plus
+// the #97 observer, and returns the brain dir so the test can read the file.
+func govServer(t *testing.T) (*mcp.Server, string, *govSink, *govNotifier) {
+	t.Helper()
+	dir := newBrainRepo(t)
+	cfg, err := classification.Load(dir)
+	require.NoError(t, err)
+	svc := capture.NewService(brainstore.New(dir), capFakeTracker{}, cfg, audit.NewSlogSink(nil))
+
+	srv := mcp.NewServer(dir, nil, nil, nil)
+	srv.WithCapture(svc, nil, capStaticTok, "local-cli", capturehttp.NewOriginResolver(nil))
+	sink, notifier := &govSink{}, &govNotifier{}
+	srv.WithGovernanceObserver(sink, notifier)
+	return srv, dir, sink, notifier
+}
+
+// authCall is call() with an Authorization header, so the static principal is
+// present and brain_edit_root's gate opens.
+func authCall(t *testing.T, srv *mcp.Server, authz, name string, args map[string]any) map[string]any {
+	t.Helper()
+	req := httptest.NewRequest(http.MethodPost, "/mcp", body(t, map[string]any{
+		"jsonrpc": "2.0", "id": 1, "method": "tools/call",
+		"params": map[string]any{"name": name, "arguments": args},
+	}))
+	if authz != "" {
+		req.Header.Set("Authorization", authz)
+	}
+	rr := httptest.NewRecorder()
+	srv.ServeHTTP(rr, req)
+	var resp map[string]any
+	require.NoError(t, json.Unmarshal(rr.Body.Bytes(), &resp))
+	return resp
+}
+
+func sha256hex(s string) string {
+	sum := sha256.Sum256([]byte(s))
+	return hex.EncodeToString(sum[:])
+}
+
+func TestBrainEditRoot_RecordsAnAuditEntryNamingPrincipalPathAndBothHashes(t *testing.T) {
+	// The gate proves a rewrite was ALLOWED. Only a record proves one HAPPENED
+	// — and brain-sync will commit it within 15 minutes authored `brain-sync`,
+	// so git history cannot tell it apart from routine note traffic.
+	srv, dir, sink, _ := govServer(t)
+
+	resp := authCall(t, srv, "Bearer "+capStaticTok, "brain_edit_root", map[string]any{
+		"path": "CLAUDE.md", "old_str": "# rules", "new_str": "# rewritten",
+	})
+	require.NotContains(t, resp, "error", "%v", resp["error"])
+
+	require.Len(t, sink.entries, 1, "exactly one audit record per governance rewrite")
+	e := sink.entries[0]
+	assert.Equal(t, "local-cli", e.Principal)
+	assert.Contains(t, e.Items, "CLAUDE.md", "the record must name what was rewritten")
+
+	joined := strings.Join(e.SecurityEvents, " ")
+	assert.Contains(t, joined, sha256hex("# rules\n"), "before-hash must be recorded")
+	assert.Contains(t, joined, sha256hex("# rewritten\n"), "after-hash must be recorded")
+	assert.NotZero(t, e.Timestamp)
+
+	got, err := os.ReadFile(filepath.Join(dir, "CLAUDE.md"))
+	require.NoError(t, err)
+	assert.Equal(t, "# rewritten\n", string(got))
+}
+
+func TestBrainEditRoot_NotifiesWithPathAndHashesButNeverContent(t *testing.T) {
+	srv, _, _, notifier := govServer(t)
+
+	resp := authCall(t, srv, "Bearer "+capStaticTok, "brain_edit_root", map[string]any{
+		"path": "CLAUDE.md", "old_str": "# rules", "new_str": "# rewritten",
+	})
+	require.NotContains(t, resp, "error", "%v", resp["error"])
+
+	require.Len(t, notifier.msgs, 1)
+	msg := notifier.msgs[0]
+	assert.Contains(t, msg, "CLAUDE.md")
+	assert.Contains(t, msg, "local-cli")
+	assert.Contains(t, msg, sha256hex("# rules\n"))
+	assert.Contains(t, msg, sha256hex("# rewritten\n"))
+
+	// The notification travels to a phone through a third-party topic. A hash
+	// identifies a rewrite; the bytes would leak whatever was written.
+	assert.NotContains(t, msg, "# rewritten", "no file content in the notification body")
+	assert.NotContains(t, msg, "# rules")
+}
+
+func TestBrainEditRoot_RefusedProducesNoRecordNoNotificationNoWrite(t *testing.T) {
+	srv, dir, sink, notifier := govServer(t)
+
+	// No Authorization header → no static principal → the gate stays shut.
+	resp := authCall(t, srv, "", "brain_edit_root", map[string]any{
+		"path": "CLAUDE.md", "old_str": "# rules", "new_str": "# rewritten",
+	})
+	require.Contains(t, resp, "error")
+
+	assert.Empty(t, sink.entries, "a refusal is not a rewrite — recording one would cry wolf")
+	assert.Empty(t, notifier.msgs)
+	got, err := os.ReadFile(filepath.Join(dir, "CLAUDE.md"))
+	require.NoError(t, err)
+	assert.Equal(t, "# rules\n", string(got))
+}
+
+func TestBrainEditRoot_FailedEditProducesNoRecordNoNotification(t *testing.T) {
+	// Authorised but no match: the door opened and nothing went through it.
+	srv, _, sink, notifier := govServer(t)
+
+	resp := authCall(t, srv, "Bearer "+capStaticTok, "brain_edit_root", map[string]any{
+		"path": "CLAUDE.md", "old_str": "absent-from-the-file", "new_str": "x",
+	})
+	require.Contains(t, resp, "error")
+
+	assert.Empty(t, sink.entries)
+	assert.Empty(t, notifier.msgs)
+}
+
+func TestBrainEdit_OnOrdinaryNotesIsNotAudited(t *testing.T) {
+	// The decided asymmetry (#97): note edits are covered by git history, which
+	// carries the same before/after bytes with a diff. Governance files get the
+	// extra trail because they change what later agents obey, not because they
+	// are harder to reconstruct.
+	srv, _, sink, notifier := govServer(t)
+
+	resp := authCall(t, srv, "Bearer "+capStaticTok, "brain_edit", map[string]any{
+		"path": "wiki/homelab/facts/n.md", "old_str": "beta", "new_str": "BETA",
+	})
+	require.NotContains(t, resp, "error", "%v", resp["error"])
+
+	assert.Empty(t, sink.entries, "note edits must not fill the governance audit trail")
+	assert.Empty(t, notifier.msgs)
+}

ingestion/internal/notes/edit.go

@@ -0,0 +1,282 @@
+package notes
+
+import (
+	"crypto/sha256"
+	"encoding/hex"
+	"errors"
+	"fmt"
+	"os"
+	"path/filepath"
+	"strings"
+	"sync"
+	"syscall"
+	"unicode/utf8"
+)
+
+// maxNoteBytes caps what Read will hand back. Generous for knowledge/ and
+// wiki/ (the largest note today is far under it) and small enough that a
+// pathological file cannot flood a caller's context.
+const maxNoteBytes = 256 << 10
+
+// Errors callers distinguish.
+var (
+	ErrNoMatch    = errors.New("notes: old_str not found")
+	ErrAmbiguous  = errors.New("notes: old_str matches more than once")
+	ErrNotUTF8    = errors.New("notes: content is not valid UTF-8")
+	ErrGitBusy    = errors.New("notes: a git operation is in progress")
+	ErrTooLarge   = errors.New("notes: file exceeds the readable size limit")
+	ErrDegenerate = errors.New("notes: edit is degenerate")
+	ErrInConflict = errors.New("notes: file contains unresolved conflict markers")
+)
+
+// pathLocks serialises edits per file within this process. The flock below
+// covers the cross-process case (two pods share the hostPath during a rolling
+// update, and brain-fix/brain-enrich write the same tree).
+var pathLocks sync.Map // string -> *sync.Mutex
+
+// Result is what a successful edit reports.
+type Result struct {
+	Path string `json:"path"`
+	// PrevContentHash is the sha256 of the file as it was before this edit.
+	// Paired with ContentHash it identifies a rewrite in an audit trail or an
+	// alert without carrying any of the bytes (hyperguild#97).
+	PrevContentHash string `json:"prev_content_hash"`
+	ContentHash     string `json:"content_hash"`
+	BytesBefore     int    `json:"bytes_before"`
+	BytesAfter      int    `json:"bytes_after"`
+	ReindexError    string `json:"reindex_error,omitempty"`
+}
+
+// Read returns a note's bytes verbatim.
+//
+// No frontmatter parsing, no excerpting, no normalisation: the entire purpose
+// is to give a caller an exact basis for an edit. An excerpt from brain_query
+// would be truncated and normalised, and every old_str built from one would
+// miss.
+func Read(brainDir, path string) ([]byte, error) {
+	t, err := Resolve(brainDir, path, ScopeNotes)
+	if err != nil {
+		return nil, err
+	}
+	return readContained(t)
+}
+
+// ReadScoped is Read for a caller already holding the privileged scope.
+func ReadScoped(brainDir, path string, scope Scope) ([]byte, error) {
+	t, err := Resolve(brainDir, path, scope)
+	if err != nil {
+		return nil, err
+	}
+	return readContained(t)
+}
+
+func readContained(t Target) ([]byte, error) {
+	fi, err := os.Stat(t.Abs)
+	if err != nil {
+		return nil, fmt.Errorf("notes: stat %s: %w", t.Rel, err)
+	}
+	if fi.Size() > maxNoteBytes {
+		return nil, fmt.Errorf("%w: %s is %d bytes (limit %d)", ErrTooLarge, t.Rel, fi.Size(), maxNoteBytes)
+	}
+	b, err := os.ReadFile(t.Abs)
+	if err != nil {
+		return nil, fmt.Errorf("notes: read %s: %w", t.Rel, err)
+	}
+	// The MCP response marshals this into a JSON string, and encoding/json
+	// substitutes U+FFFD for invalid UTF-8 — silently returning different
+	// bytes than the file holds, which defeats "verbatim" entirely.
+	if !utf8.Valid(b) {
+		return nil, fmt.Errorf("%w: %s", ErrNotUTF8, t.Rel)
+	}
+	return b, nil
+}
+
+// Edit replaces exactly one occurrence of oldStr with newStr, in place.
+//
+// Fails loudly and writes nothing on: zero matches, more than one match
+// (counted with overlap), degenerate arguments, invalid UTF-8, a git operation
+// in flight, or unresolved conflict markers in the target.
+func Edit(brainDir, path, oldStr, newStr string, scope Scope) (Result, error) {
+	t, err := Resolve(brainDir, path, scope)
+	if err != nil {
+		return Result{}, err
+	}
+	if oldStr == "" {
+		return Result{}, fmt.Errorf("%w: old_str is empty", ErrDegenerate)
+	}
+	if oldStr == newStr {
+		return Result{}, fmt.Errorf("%w: old_str equals new_str", ErrDegenerate)
+	}
+	if !utf8.ValidString(newStr) {
+		return Result{}, fmt.Errorf("%w: new_str", ErrNotUTF8)
+	}
+	// brain#2 guard: refuse to write to a tree the sync does not commit from.
+	if err := assertSyncSource(brainDir); err != nil {
+		return Result{}, err
+	}
+	if err := assertGitIdle(brainDir); err != nil {
+		return Result{}, err
+	}
+
+	mu := lockFor(t.Abs)
+	mu.Lock()
+	defer mu.Unlock()
+
+	unlock, err := flockFile(t.Abs)
+	if err != nil {
+		return Result{}, err
+	}
+	defer unlock()
+
+	// Read INSIDE the lock: the match count must be evaluated against the
+	// bytes we are about to overwrite, or a concurrent edit is silently lost
+	// while both callers are told they succeeded.
+	before, err := readContained(t)
+	if err != nil {
+		return Result{}, err
+	}
+	if hasConflictMarkers(before) {
+		return Result{}, fmt.Errorf("%w: %s", ErrInConflict, t.Rel)
+	}
+
+	switch n := countOverlapping(string(before), oldStr); {
+	case n == 0:
+		return Result{}, fmt.Errorf("%w in %s (0 matches)%s", ErrNoMatch, t.Rel, crlfHint(string(before), oldStr))
+	case n > 1:
+		return Result{}, fmt.Errorf("%w in %s (%d matches) — extend old_str until it is unique",
+			ErrAmbiguous, t.Rel, n)
+	}
+
+	after := strings.Replace(string(before), oldStr, newStr, 1)
+	if err := writeAtomic(t.Abs, []byte(after)); err != nil {
+		return Result{}, err
+	}
+
+	sum := sha256.Sum256([]byte(after))
+	prev := sha256.Sum256(before)
+	return Result{
+		Path:            t.Rel,
+		PrevContentHash: hex.EncodeToString(prev[:]),
+		ContentHash:     hex.EncodeToString(sum[:]),
+		BytesBefore:     len(before),
+		BytesAfter:      len(after),
+	}, nil
+}
+
+// countOverlapping counts occurrences allowing overlap. bytes.Count does not:
+// "aa" occurs once in "aaa" by its reckoning, so a Count-based uniqueness
+// check accepts an ambiguous edit as unique.
+func countOverlapping(s, sub string) int {
+	var n, from int
+	for {
+		i := strings.Index(s[from:], sub)
+		if i < 0 {
+			return n
+		}
+		n++
+		from += i + 1
+	}
+}
+
+// crlfHint reports whether the caller's old_str would have matched if line
+// endings agreed. It never normalises — that would break byte-identity — but
+// an opaque "no match" is unfixable from the caller's side, and a stripped \r
+// is the likeliest cause.
+func crlfHint(content, oldStr string) string {
+	if strings.Contains(content, "\r\n") && !strings.Contains(oldStr, "\r\n") {
+		if strings.Contains(strings.ReplaceAll(content, "\r\n", "\n"), oldStr) {
+			return " — the file uses CRLF line endings and old_str does not; include the \\r"
+		}
+	}
+	return ""
+}
+
+func hasConflictMarkers(b []byte) bool {
+	s := string(b)
+	return strings.Contains(s, "\n<<<<<<< ") || strings.HasPrefix(s, "<<<<<<< ")
+}
+
+// assertGitIdle refuses while brain-sync (or brain-fix, or a human) has a git
+// operation in flight. An edit landing between `git merge` and its abort is
+// either hard-reset away or committed as a wrongly resolved conflict — and the
+// caller was told it succeeded.
+func assertGitIdle(brainDir string) error {
+	for _, marker := range []string{"MERGE_HEAD", "REBASE_HEAD", "CHERRY_PICK_HEAD", "index.lock"} {
+		if _, err := os.Stat(filepath.Join(brainDir, ".git", marker)); err == nil {
+			return fmt.Errorf("%w (.git/%s present) — retry after the sync settles", ErrGitBusy, marker)
+		}
+	}
+	return nil
+}
+
+func lockFor(abs string) *sync.Mutex {
+	v, _ := pathLocks.LoadOrStore(abs, &sync.Mutex{})
+	return v.(*sync.Mutex)
+}
+
+// flockFile takes an advisory lock covering other processes sharing the
+// hostPath — during a RollingUpdate two ingestion pods do.
+func flockFile(abs string) (func(), error) {
+	f, err := os.OpenFile(abs+".lock", os.O_CREATE|os.O_RDWR, 0o644)
+	if err != nil {
+		return nil, fmt.Errorf("notes: lock %s: %w", filepath.Base(abs), err)
+	}
+	if err := syscall.Flock(int(f.Fd()), syscall.LOCK_EX); err != nil {
+		_ = f.Close()
+		return nil, fmt.Errorf("notes: flock %s: %w", filepath.Base(abs), err)
+	}
+	return func() {
+		_ = syscall.Flock(int(f.Fd()), syscall.LOCK_UN)
+		_ = f.Close()
+		_ = os.Remove(abs + ".lock")
+	}, nil
+}
+
+// writeAtomic replaces the file via a same-directory temp + rename, then
+// fsyncs the directory so the rename survives a node crash.
+//
+// Same directory because a rename across the hostPath boundary fails with
+// EXDEV; the ".tmp" suffix because the tree's .gitignore already covers it, so
+// a leaked temp is not committed by `git add -A`.
+func writeAtomic(abs string, content []byte) error {
+	dir := filepath.Dir(abs)
+	tmp, err := os.CreateTemp(dir, filepath.Base(abs)+".*.tmp")
+	if err != nil {
+		return fmt.Errorf("notes: temp file: %w", err)
+	}
+	tmpName := tmp.Name()
+	defer func() { _ = os.Remove(tmpName) }() // no-op once renamed
+
+	if _, err := tmp.Write(content); err != nil {
+		_ = tmp.Close()
+		return fmt.Errorf("notes: write temp: %w", err)
+	}
+	if err := tmp.Sync(); err != nil {
+		_ = tmp.Close()
+		return fmt.Errorf("notes: sync temp: %w", err)
+	}
+	if err := tmp.Close(); err != nil {
+		return fmt.Errorf("notes: close temp: %w", err)
+	}
+	// CreateTemp makes 0600; every other writer of this tree uses 0644 and
+	// other uids read it.
+	if err := os.Chmod(tmpName, 0o644); err != nil {
+		return fmt.Errorf("notes: chmod temp: %w", err)
+	}
+	if err := os.Rename(tmpName, abs); err != nil {
+		return fmt.Errorf("notes: rename: %w", err)
+	}
+	return fsyncDir(dir)
+}
+
+func fsyncDir(dir string) error {
+	d, err := os.Open(dir)
+	if err != nil {
+		return fmt.Errorf("notes: open dir: %w", err)
+	}
+	defer func() { _ = d.Close() }()
+	if err := d.Sync(); err != nil {
+		return fmt.Errorf("notes: fsync dir: %w", err)
+	}
+	return nil
+}

ingestion/internal/notes/edit_test.go

@@ -0,0 +1,256 @@
+package notes
+
+import (
+	"os"
+	"path/filepath"
+	"strings"
+	"sync"
+	"testing"
+)
+
+func seed(t *testing.T, dir, rel, content string) string {
+	t.Helper()
+	abs := filepath.Join(dir, rel)
+	if err := os.MkdirAll(filepath.Dir(abs), 0o755); err != nil {
+		t.Fatal(err)
+	}
+	if err := os.WriteFile(abs, []byte(content), 0o644); err != nil {
+		t.Fatal(err)
+	}
+	return abs
+}
+
+func read(t *testing.T, abs string) string {
+	t.Helper()
+	b, err := os.ReadFile(abs)
+	if err != nil {
+		t.Fatal(err)
+	}
+	return string(b)
+}
+
+func TestRead_ReturnsVerbatimBytes(t *testing.T) {
+	dir := newBrain(t)
+	const body = "---\ntitle: 'x'\n---\n\nbody with  double  spaces\r\nand a CRLF line\nno trailing newline"
+	seed(t, dir, "knowledge/verbatim.md", body)
+
+	got, err := Read(dir, "knowledge/verbatim.md")
+	if err != nil {
+		t.Fatalf("Read: %v", err)
+	}
+	if string(got) != body {
+		t.Errorf("Read returned %q, want the file byte-for-byte", got)
+	}
+}
+
+func TestRead_RefusesNonUTF8RatherThanCorruptingIt(t *testing.T) {
+	// The response is marshalled into a JSON string, and encoding/json
+	// replaces invalid UTF-8 with U+FFFD. Returning it would mean every
+	// old_str the caller derives from that region can never match, and
+	// pasting it back would write the corruption to disk permanently.
+	dir := newBrain(t)
+	seed(t, dir, "knowledge/bad.md", "ok \xff\xfe broken\n")
+
+	if _, err := Read(dir, "knowledge/bad.md"); err == nil {
+		t.Fatal("Read accepted invalid UTF-8; it must refuse rather than silently substitute")
+	}
+}
+
+func TestEdit_ReplacesExactlyOnceAndLeavesEveryOtherByteAlone(t *testing.T) {
+	dir := newBrain(t)
+	const body = "---\ntitle: 'x'\n---\n\nalpha\nbeta\ngamma\n"
+	abs := seed(t, dir, "knowledge/n.md", body)
+
+	res, err := Edit(dir, "knowledge/n.md", "beta", "BETA", ScopeNotes)
+	if err != nil {
+		t.Fatalf("Edit: %v", err)
+	}
+	want := strings.Replace(body, "beta", "BETA", 1)
+	if got := read(t, abs); got != want {
+		t.Errorf("file = %q, want %q", got, want)
+	}
+	if res.ContentHash == "" {
+		t.Error("result should carry a content hash for read-after-write confirmation")
+	}
+}
+
+func TestEdit_RefusesAmbiguousAndAbsentMatchesWithoutWriting(t *testing.T) {
+	dir := newBrain(t)
+	const body = "alpha\nalpha\n"
+	abs := seed(t, dir, "knowledge/n.md", body)
+
+	cases := []struct{ name, old, want string }{
+		{"two matches", "alpha", "2"},
+		{"no match", "delta", "0"},
+	}
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			_, err := Edit(dir, "knowledge/n.md", tc.old, "x", ScopeNotes)
+			if err == nil {
+				t.Fatal("want a loud refusal")
+			}
+			if !strings.Contains(err.Error(), tc.want) {
+				t.Errorf("error must state the match count, got %v", err)
+			}
+			if got := read(t, abs); got != body {
+				t.Errorf("file changed on a refused edit: %q", got)
+			}
+		})
+	}
+}
+
+func TestEdit_CountsOverlappingMatches(t *testing.T) {
+	// bytes.Count counts NON-overlapping occurrences: "aa" in "aaa" is 1, so
+	// a Count-based uniqueness check calls an ambiguous edit unique.
+	dir := newBrain(t)
+	seed(t, dir, "knowledge/n.md", "aaa\n")
+
+	if _, err := Edit(dir, "knowledge/n.md", "aa", "b", ScopeNotes); err == nil {
+		t.Fatal("overlapping matches must be reported as ambiguous")
+	}
+}
+
+func TestEdit_RejectsDegenerateArguments(t *testing.T) {
+	dir := newBrain(t)
+	seed(t, dir, "knowledge/n.md", "alpha\n")
+
+	if _, err := Edit(dir, "knowledge/n.md", "", "x", ScopeNotes); err == nil {
+		t.Error("empty old_str must be rejected")
+	}
+	if _, err := Edit(dir, "knowledge/n.md", "alpha", "alpha", ScopeNotes); err == nil {
+		t.Error("a no-op edit must be rejected rather than burning a sync cycle")
+	}
+}
+
+func TestEdit_DeletionIsLegal(t *testing.T) {
+	dir := newBrain(t)
+	abs := seed(t, dir, "knowledge/n.md", "keep\ndrop\n")
+
+	if _, err := Edit(dir, "knowledge/n.md", "drop\n", "", ScopeNotes); err != nil {
+		t.Fatalf("Edit: %v", err)
+	}
+	if got := read(t, abs); got != "keep\n" {
+		t.Errorf("file = %q, want %q", got, "keep\n")
+	}
+}
+
+func TestEdit_ReportsWhenOnlyLineEndingsDiffer(t *testing.T) {
+	// The likeliest real failure: the caller's old_str lost its \r. Never
+	// normalise silently — that breaks byte-identity — but say so, because an
+	// opaque "no match" is unfixable from the caller's side.
+	dir := newBrain(t)
+	seed(t, dir, "knowledge/n.md", "alpha\r\nbeta\r\n")
+
+	_, err := Edit(dir, "knowledge/n.md", "alpha\nbeta", "x", ScopeNotes)
+	if err == nil {
+		t.Fatal("want a refusal")
+	}
+	if !strings.Contains(err.Error(), "line ending") {
+		t.Errorf("error should point at CRLF, got %v", err)
+	}
+}
+
+func TestEdit_RefusesWhileGitIsMidOperation(t *testing.T) {
+	// brain-sync runs `git merge -X ours origin/main`, and aborts on conflict
+	// with a hard reset. An edit landing inside that window is either erased
+	// or committed as a wrongly-resolved conflict, and the caller was told it
+	// succeeded.
+	dir := newBrain(t)
+	seed(t, dir, "knowledge/n.md", "alpha\n")
+
+	for _, marker := range []string{"MERGE_HEAD", "index.lock", "REBASE_HEAD"} {
+		t.Run(marker, func(t *testing.T) {
+			p := filepath.Join(dir, ".git", marker)
+			if err := os.WriteFile(p, []byte("x"), 0o644); err != nil {
+				t.Fatal(err)
+			}
+			defer func() { _ = os.Remove(p) }()
+
+			if _, err := Edit(dir, "knowledge/n.md", "alpha", "beta", ScopeNotes); err == nil {
+				t.Errorf("edit accepted while %s exists", marker)
+			}
+		})
+	}
+}
+
+func TestEdit_RefusesAFileHoldingConflictMarkers(t *testing.T) {
+	dir := newBrain(t)
+	seed(t, dir, "knowledge/n.md", "<<<<<<< HEAD\nours\n=======\ntheirs\n>>>>>>> origin/main\n")
+
+	if _, err := Edit(dir, "knowledge/n.md", "ours", "x", ScopeNotes); err == nil {
+		t.Fatal("editing one side of an unresolved conflict must be refused")
+	}
+}
+
+func TestEdit_ConcurrentEditsBothApply(t *testing.T) {
+	// Read-modify-write with no lock loses one update AND reports success for
+	// both. old_str uniqueness does not protect against it: uniqueness is
+	// evaluated against the stale read.
+	dir := newBrain(t)
+	abs := seed(t, dir, "knowledge/n.md", "alpha\nbeta\n")
+
+	var wg sync.WaitGroup
+	errs := make([]error, 2)
+	for i, pair := range [][2]string{{"alpha", "ALPHA"}, {"beta", "BETA"}} {
+		wg.Add(1)
+		go func(i int, old, new string) {
+			defer wg.Done()
+			_, errs[i] = Edit(dir, "knowledge/n.md", old, new, ScopeNotes)
+		}(i, pair[0], pair[1])
+	}
+	wg.Wait()
+
+	for i, err := range errs {
+		if err != nil {
+			t.Fatalf("edit %d failed: %v", i, err)
+		}
+	}
+	if got := read(t, abs); got != "ALPHA\nBETA\n" {
+		t.Errorf("file = %q, want both edits applied", got)
+	}
+}
+
+func TestEdit_LeavesNoTemporaryFileBehind(t *testing.T) {
+	dir := newBrain(t)
+	seed(t, dir, "knowledge/n.md", "alpha\n")
+
+	if _, err := Edit(dir, "knowledge/n.md", "alpha", "beta", ScopeNotes); err != nil {
+		t.Fatal(err)
+	}
+	entries, err := os.ReadDir(filepath.Join(dir, "knowledge"))
+	if err != nil {
+		t.Fatal(err)
+	}
+	for _, e := range entries {
+		if e.Name() != "note.md" && e.Name() != "n.md" {
+			t.Errorf("stray file left in the tree: %q (git add -A would commit it)", e.Name())
+		}
+	}
+}
+
+func TestEdit_PreservesReadableMode(t *testing.T) {
+	// Other uids write this tree (brain-fix, brain-enrich). os.CreateTemp
+	// yields 0600 and rename preserves it, which would break them silently.
+	dir := newBrain(t)
+	abs := seed(t, dir, "knowledge/n.md", "alpha\n")
+
+	if _, err := Edit(dir, "knowledge/n.md", "alpha", "beta", ScopeNotes); err != nil {
+		t.Fatal(err)
+	}
+	fi, err := os.Stat(abs)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if fi.Mode().Perm() != 0o644 {
+		t.Errorf("mode = %v, want 0644", fi.Mode().Perm())
+	}
+}
+
+func TestEdit_RefusesNonUTF8Replacement(t *testing.T) {
+	dir := newBrain(t)
+	seed(t, dir, "knowledge/n.md", "alpha\n")
+
+	if _, err := Edit(dir, "knowledge/n.md", "alpha", "bad \xff", ScopeNotes); err == nil {
+		t.Fatal("new_str with invalid UTF-8 must be refused, not written")
+	}
+}

ingestion/internal/notes/path.go

@@ -0,0 +1,201 @@
+// Package notes provides containment-checked read and surgical edit of brain
+// notes on the live tree that brain-sync commits from.
+//
+// Everything here is defensive for one reason: brain content is ingested from
+// the outside world, that content reaches a model's context via brain_query /
+// brain_context, and these primitives are callable by that model. A path bug
+// is therefore not a path bug — it is an outside author choosing a file.
+package notes
+
+import (
+	"errors"
+	"fmt"
+	"os"
+	"path/filepath"
+	"strings"
+)
+
+// Scope is which part of the tree a caller is allowed to touch.
+type Scope int
+
+const (
+	// ScopeNotes is the model-facing scope: markdown under knowledge/ and
+	// wiki/ only.
+	ScopeNotes Scope = iota
+	// ScopeRoot admits the named governance files and NOTHING else. It exists so
+	// there is an MCP path to CLAUDE.md at all — the absence of one is what
+	// pushed a session to edit via the gitea REST API on 2026-06-16, which
+	// never reached the live tree and wedged brain-sync for hours. It is NOT
+	// reachable from the model-facing toolset; see tools_notes.go.
+	ScopeRoot
+)
+
+// rootFiles is an exact-name allowlist, deliberately not a "root *.md" glob.
+//
+// Two reasons. log.md is gitignored, so an edit to it applies to the live tree
+// and never reaches gitea — the exact shape of brain#2, admitted by a glob. And
+// a glob auto-admits whatever governance file appears next (AGENTS.md, a
+// bootstrap doc) with nobody reviewing that decision.
+var rootFiles = map[string]bool{"CLAUDE.md": true}
+
+// noteDirs are the two trees that hold notes.
+var noteDirs = []string{"knowledge/", "wiki/"}
+
+// deniedPrefixes are refused in every scope, independent of the allowlist.
+//
+// A static backstop rather than a consequence of the allowlist being right:
+// .git/hooks/post-commit is executed by brain-sync inside a pod holding the
+// Gitea push credentials, and archive/ + sessions/ hold verbatim session
+// transcripts that BM25 never indexes and so nobody thinks of as reachable.
+var deniedPrefixes = []string{".git/", "archive/", "sessions/", "raw/", "training-data/", ".obsidian/"}
+
+// Errors are sentinels so callers can distinguish a refusal from an I/O fault.
+var (
+	ErrOutsideBrain  = errors.New("notes: path escapes the brain dir")
+	ErrNotAllowed    = errors.New("notes: path is not editable through this tool")
+	ErrNotSyncSource = errors.New("notes: brain dir is not the sync source")
+)
+
+// Target is a resolved, contained path.
+type Target struct {
+	Rel string // brain-relative, cleaned
+	Abs string // absolute, symlink-checked
+}
+
+// Resolve canonicalises path, then checks it against the scope's allowlist,
+// then re-checks containment after symlink evaluation.
+//
+// Order matters and is the point: a prefix test on the raw argument accepts
+// "knowledge/../.git/hooks/post-commit", because Clean only collapses it
+// afterwards.
+func Resolve(brainDir, path string, scope Scope) (Target, error) {
+	// Note: the sync-source assertion lives on the WRITE path (Edit), not
+	// here. brain#2 was an edit landing where the sync could not commit it; a
+	// read from such a tree is merely a read.
+
+	// 1. Canonicalise. Rooting at "/" makes Clean collapse every ".." that
+	//    would otherwise escape, and ToSlash normalises separators first.
+	clean := strings.TrimPrefix(filepath.Clean("/"+filepath.ToSlash(path)), "/")
+	if clean == "" || clean == "." {
+		return Target{}, fmt.Errorf("%w: empty path", ErrNotAllowed)
+	}
+
+	// 2. Static denials, before any allowlist reasoning.
+	for _, d := range deniedPrefixes {
+		if strings.HasPrefix(clean+"/", d) {
+			return Target{}, fmt.Errorf("%w: %q is not note content", ErrNotAllowed, clean)
+		}
+	}
+	if strings.HasPrefix(filepath.Base(clean), ".") {
+		return Target{}, fmt.Errorf("%w: dotfiles are not editable", ErrNotAllowed)
+	}
+
+	// 3. Scope allowlist.
+	if err := allow(clean, scope); err != nil {
+		return Target{}, err
+	}
+
+	// 4. Containment, lexically.
+	abs := filepath.Join(brainDir, filepath.FromSlash(clean))
+	if !within(brainDir, abs) {
+		return Target{}, fmt.Errorf("%w: %q", ErrOutsideBrain, path)
+	}
+
+	// 5. Containment again, after resolving symlinks. The tree is Syncthing
+	//    managed, so a link can appear without anyone adding one on purpose;
+	//    a symlinked note or directory would otherwise read and write outside
+	//    the brain entirely.
+	if err := assertNoSymlinkEscape(brainDir, abs); err != nil {
+		return Target{}, err
+	}
+
+	return Target{Rel: clean, Abs: abs}, nil
+}
+
+// allow enforces the scope's allowlist. The two scopes are DISJOINT, not
+// nested: the privileged root scope reaches governance files and nothing else,
+// so the tool that carries the higher privilege carries the smaller reach.
+func allow(clean string, scope Scope) error {
+	if scope == ScopeRoot {
+		if !rootFiles[clean] {
+			return fmt.Errorf("%w: %q is not a governance file", ErrNotAllowed, clean)
+		}
+		return nil
+	}
+	if !isNotePath(clean) {
+		return fmt.Errorf("%w: %q is outside knowledge/ and wiki/", ErrNotAllowed, clean)
+	}
+	return nil
+}
+
+func isNotePath(clean string) bool {
+	var inDir bool
+	for _, d := range noteDirs {
+		if strings.HasPrefix(clean, d) {
+			inDir = true
+			break
+		}
+	}
+	if !inDir || filepath.Ext(clean) != ".md" {
+		return false
+	}
+	// _index.md is generated per wing, and brain-sync's conflict resolver
+	// deliberately resolves it in origin's favour — an edit here is reverted
+	// by design, so accepting one would be a lie.
+	return filepath.Base(clean) != "_index.md"
+}
+
+func within(root, abs string) bool {
+	rel, err := filepath.Rel(root, abs)
+	if err != nil {
+		return false
+	}
+	return rel != ".." && !strings.HasPrefix(rel, ".."+string(filepath.Separator))
+}
+
+// assertNoSymlinkEscape resolves the deepest existing ancestor of abs and
+// requires it to still be inside brainDir. The file itself need not exist
+// (a create is legal); its directory chain must be real and contained.
+func assertNoSymlinkEscape(brainDir, abs string) error {
+	if fi, err := os.Lstat(abs); err == nil && fi.Mode()&os.ModeSymlink != 0 {
+		return fmt.Errorf("%w: %q is a symlink", ErrNotAllowed, abs)
+	}
+
+	realRoot, err := filepath.EvalSymlinks(brainDir)
+	if err != nil {
+		return fmt.Errorf("notes: resolve brain dir: %w", err)
+	}
+
+	dir := filepath.Dir(abs)
+	for {
+		resolved, err := filepath.EvalSymlinks(dir)
+		if err == nil {
+			if !within(realRoot, resolved) {
+				return fmt.Errorf("%w: %q resolves outside the brain dir", ErrOutsideBrain, dir)
+			}
+			return nil
+		}
+		if !os.IsNotExist(err) {
+			return fmt.Errorf("notes: resolve %q: %w", dir, err)
+		}
+		parent := filepath.Dir(dir)
+		if parent == dir {
+			return fmt.Errorf("%w: no existing ancestor for %q", ErrOutsideBrain, abs)
+		}
+		dir = parent
+	}
+}
+
+// assertSyncSource refuses a brain dir that is not a git working tree.
+//
+// brain#2 was an edit landing on a directory the sync does not commit from.
+// That cannot be prevented by review — a wrong INGEST_BRAIN_DIR in a future
+// manifest reintroduces it while every unit test still passes — so it is
+// asserted at runtime, the same check brain-sync itself makes before running.
+func assertSyncSource(brainDir string) error {
+	if fi, err := os.Stat(filepath.Join(brainDir, ".git")); err != nil || !fi.IsDir() {
+		return fmt.Errorf("%w: %q has no .git, so an edit there would never be committed",
+			ErrNotSyncSource, brainDir)
+	}
+	return nil
+}

ingestion/internal/notes/path_test.go

@@ -0,0 +1,146 @@
+package notes
+
+import (
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+)
+
+// newBrain lays out a tree shaped like the live one: a git dir, note dirs,
+// and the non-note areas that must never be reachable.
+func newBrain(t *testing.T) string {
+	t.Helper()
+	dir := t.TempDir()
+	for _, d := range []string{
+		".git/hooks", "knowledge", "wiki/homelab/facts", "archive/claude-sessions/koala",
+		"sessions", "raw", ".obsidian",
+	} {
+		if err := os.MkdirAll(filepath.Join(dir, d), 0o755); err != nil {
+			t.Fatal(err)
+		}
+	}
+	write := func(p, s string) {
+		t.Helper()
+		if err := os.WriteFile(filepath.Join(dir, p), []byte(s), 0o644); err != nil {
+			t.Fatal(err)
+		}
+	}
+	write("CLAUDE.md", "# instructions\n")
+	write("log.md", "ignored scratch\n")
+	write("knowledge/note.md", "# note\n")
+	write("wiki/homelab/facts/f.md", "# fact\n")
+	write("wiki/homelab/_index.md", "# index\n")
+	write("archive/claude-sessions/koala/session-1.md", "verbatim transcript\n")
+	write("sessions/s.jsonl", "{}\n")
+	write(".git/hooks/post-commit", "#!/bin/sh\n")
+	return dir
+}
+
+func TestResolve_AcceptsNotePaths(t *testing.T) {
+	dir := newBrain(t)
+	for _, p := range []string{"knowledge/note.md", "wiki/homelab/facts/f.md"} {
+		got, err := Resolve(dir, p, ScopeNotes)
+		if err != nil {
+			t.Errorf("Resolve(%q): %v", p, err)
+			continue
+		}
+		if !strings.HasPrefix(got.Abs, dir) {
+			t.Errorf("Resolve(%q).Abs = %q, want inside %q", p, got.Abs, dir)
+		}
+	}
+}
+
+func TestResolve_RefusesEscapesAndNonNoteAreas(t *testing.T) {
+	dir := newBrain(t)
+
+	// Canonicalisation must happen BEFORE the allowlist check. A prefix test
+	// on the raw argument accepts every one of these: "knowledge/..%2f" style
+	// inputs collapse to .git, and .git/hooks/post-commit is executed by
+	// brain-sync inside a pod holding the Gitea push credentials.
+	refuse := []string{
+		"knowledge/../.git/hooks/post-commit",
+		"wiki/../.git/config",
+		"./knowledge/../.git/config",
+		"knowledge//../.git/config",
+		"../../etc/passwd",
+		"/etc/passwd",
+		".git/config",
+		"archive/claude-sessions/koala/session-1.md", // verbatim transcripts
+		"sessions/s.jsonl",
+		"raw/x.md",
+		".obsidian/workspace.json",
+		"log.md",                 // gitignored: an edit here would never sync (brain#2)
+		"wiki/homelab/_index.md", // generated; brain-sync resolves it in origin's favour
+		"knowledge/note.txt",     // not markdown
+		"CLAUDE.md",              // root file — not reachable in the notes scope
+	}
+	for _, p := range refuse {
+		if _, err := Resolve(dir, p, ScopeNotes); err == nil {
+			t.Errorf("Resolve(%q) = nil error, want refusal", p)
+		}
+	}
+}
+
+func TestResolve_RootScopeAdmitsOnlyNamedFiles(t *testing.T) {
+	dir := newBrain(t)
+
+	if _, err := Resolve(dir, "CLAUDE.md", ScopeRoot); err != nil {
+		t.Errorf("Resolve(CLAUDE.md, ScopeRoot): %v", err)
+	}
+	// log.md is the reason this is an exact-name list and not "root *.md":
+	// it is gitignored, so an edit would apply locally and never reach gitea.
+	for _, p := range []string{"log.md", "AGENTS.md", "knowledge/note.md", ".git/config"} {
+		if _, err := Resolve(dir, p, ScopeRoot); err == nil {
+			t.Errorf("Resolve(%q, ScopeRoot) = nil error, want refusal", p)
+		}
+	}
+}
+
+func TestResolve_RefusesSymlinkedTargets(t *testing.T) {
+	dir := newBrain(t)
+	outside := filepath.Join(t.TempDir(), "secret.md")
+	if err := os.WriteFile(outside, []byte("secret\n"), 0o644); err != nil {
+		t.Fatal(err)
+	}
+	// The tree is Syncthing-managed, so a symlink can appear without anyone
+	// adding one deliberately. Containment must survive it.
+	if err := os.Symlink(outside, filepath.Join(dir, "knowledge/link.md")); err != nil {
+		t.Skipf("symlinks unavailable: %v", err)
+	}
+	if _, err := Resolve(dir, "knowledge/link.md", ScopeNotes); err == nil {
+		t.Error("Resolve followed a symlink out of the brain dir")
+	}
+
+	if err := os.Symlink(filepath.Dir(outside), filepath.Join(dir, "wiki/escape")); err != nil {
+		t.Skipf("symlinks unavailable: %v", err)
+	}
+	if _, err := Resolve(dir, "wiki/escape/secret.md", ScopeNotes); err == nil {
+		t.Error("Resolve followed a symlinked directory out of the brain dir")
+	}
+}
+
+func TestResolve_RefusesWhenTheTreeIsNotTheSyncSource(t *testing.T) {
+	// brain#2 was "wrote to a directory the sync does not commit from". No
+	// code review prevents that recurring — a wrong INGEST_BRAIN_DIR in a
+	// future manifest reintroduces it and every unit test still passes. So
+	// assert the property at runtime, the way brain-sync itself does.
+	dir := t.TempDir()
+	if err := os.MkdirAll(filepath.Join(dir, "knowledge"), 0o755); err != nil {
+		t.Fatal(err)
+	}
+	if err := os.WriteFile(filepath.Join(dir, "knowledge/note.md"), []byte("x"), 0o644); err != nil {
+		t.Fatal(err)
+	}
+	// Reads are fine anywhere; the guard belongs to writes.
+	if _, err := Resolve(dir, "knowledge/note.md", ScopeNotes); err != nil {
+		t.Fatalf("Resolve should not require .git for a read: %v", err)
+	}
+	_, err := Edit(dir, "knowledge/note.md", "x", "y", ScopeNotes)
+	if err == nil {
+		t.Fatal("Edit accepted a tree with no .git — the write would never be committed")
+	}
+	if !strings.Contains(err.Error(), "sync source") {
+		t.Errorf("error should name the cause, got %v", err)
+	}
+}

ingestion/internal/pipeline/build.go

@@ -76,6 +76,9 @@ func buildFrontmatter(rp RawPage, date string) string {
 		}
 		fmt.Fprintf(&sb, "date_ingested: %s\n", date)
 		fmt.Fprintf(&sb, "last_updated: %s\n", date)
+		if rp.SourceHash != "" {
+			fmt.Fprintf(&sb, "source_hash: %s\n", rp.SourceHash)
+		}
 		if rp.Source != "" {
 			fmt.Fprintf(&sb, "source: %s\n", yamlScalar(rp.Source))
 		}

ingestion/internal/pipeline/parse.go

@@ -20,9 +20,13 @@ type RawPage struct {
 	// ingested content's own frontmatter (see parseContentFrontmatter) — never
 	// set by the LLM. json:"-" keeps them immune to same-named keys the LLM
 	// might emit. Only meaningful for Type == "source".
-	Source    string `json:"-"`
-	Author    string `json:"-"`
-	Published string `json:"-"`
+	// SourceHash is sha256 of the ingested BODY (frontmatter excluded), stamped
+	// so the next ingest of the same source can tell "unchanged" from "edited"
+	// without re-running extraction (#105).
+	SourceHash string `json:"-"`
+	Source     string `json:"-"`
+	Author     string `json:"-"`
+	Published  string `json:"-"`
 }
 
 // ParseRawPages parses LLM output as a JSON array of RawPage objects.
@@ -112,6 +116,11 @@ type sourceMeta struct {
 	Source    string
 	Author    string
 	Published string
+	// BodyHash is sha256 of the content with any leading frontmatter removed.
+	// Body-only on purpose: a re-crawl that stamps a new fetch date has not
+	// changed the source, and treating that as drift would re-extract the
+	// whole corpus every time a crawler ran (#105).
+	BodyHash string
 }
 
 // parseContentFrontmatter extracts source/author/published from a leading
@@ -154,6 +163,7 @@ func applySourceMeta(pages []RawPage, meta sourceMeta) {
 		if pages[i].Type != "source" {
 			continue
 		}
+		pages[i].SourceHash = meta.BodyHash
 		pages[i].Source = meta.Source
 		pages[i].Author = meta.Author
 		pages[i].Published = meta.Published

ingestion/internal/pipeline/pipeline.go

@@ -3,6 +3,8 @@ package pipeline
 
 import (
 	"context"
+	"crypto/sha256"
+	"encoding/hex"
 	"fmt"
 	"os"
 	"path/filepath"
@@ -26,6 +28,11 @@ type Config struct {
 type Result struct {
 	Pages    []string // relative paths written (or would-be written in dry-run)
 	Warnings []string
+	// Skipped reports that the source was byte-identical to the last ingest,
+	// so no extraction ran and nothing was written. Reported rather than
+	// silent: a run that does nothing must say so, or "the pipeline is fine"
+	// and "the pipeline stopped working" look the same (#105, #107).
+	Skipped bool
 }
 
 // Run ingests content and writes structured wiki pages to brainDir/wiki/.
@@ -43,6 +50,18 @@ func Run(ctx context.Context, cfg Config, brainDir, content, source string, dryR
 
 	sourceSlug := wiki.Slug(source)
 	date := time.Now().UTC().Format("2006-01-02")
+
+	// Drift check BEFORE any extraction: an unchanged source costs a full LLM
+	// pass over every chunk to produce a result that merges to nothing (#105).
+	// The merge is already idempotent (#103), so this is cost and signal, not
+	// correctness — which is why it is safe for it to be an early return.
+	bodyHash := hashBody(content)
+	if !dryRun && bodyHash != "" && bodyHash == storedSourceHash(brainDir, sourceSlug) {
+		return Result{Skipped: true, Warnings: []string{
+			fmt.Sprintf("source %q unchanged since last ingest (hash %s) — extraction skipped", source, bodyHash[:12]),
+		}}, nil
+	}
+
 	chunks := Chunk(content, cfg.ChunkSize)
 
 	var allRaw []RawPage
@@ -59,7 +78,9 @@ func Run(ctx context.Context, cfg Config, brainDir, content, source string, dryR
 		allWarnings = append(allWarnings, warnings...)
 	}
 
-	applySourceMeta(allRaw, parseContentFrontmatter(content))
+	meta := parseContentFrontmatter(content)
+	meta.BodyHash = bodyHash
+	applySourceMeta(allRaw, meta)
 
 	return buildAndWrite(allRaw, sourceSlug, date, brainDir, source, inventory, allWarnings, dryRun)
 }
@@ -96,7 +117,11 @@ func buildAndWrite(rawPages []RawPage, sourceSlug, date, brainDir, source string
 			if err := os.MkdirAll(filepath.Dir(dest), 0o755); err != nil {
 				return Result{}, fmt.Errorf("mkdir for %s: %w", page.Path, err)
 			}
-			if err := os.WriteFile(dest, []byte(page.Content), 0o644); err != nil {
+			content, err := reviseOrCreate(dest, page)
+			if err != nil {
+				return Result{}, err
+			}
+			if err := os.WriteFile(dest, []byte(content), 0o644); err != nil {
 				return Result{}, fmt.Errorf("write %s: %w", page.Path, err)
 			}
 		}
@@ -115,6 +140,75 @@ func buildAndWrite(rawPages []RawPage, sourceSlug, date, brainDir, source string
 	return Result{Pages: written, Warnings: warnings}, nil
 }
 
+// hashBody returns sha256 of content with any leading frontmatter block
+// removed, or "" for empty content.
+//
+// Body-only on purpose: a re-crawl that stamps a new fetch date has not changed
+// the source, and treating that as drift would re-extract the corpus every time
+// a crawler ran (#105).
+func hashBody(content string) string {
+	body := content
+	if strings.HasPrefix(body, "---\n") {
+		if _, rest, found := strings.Cut(body[4:], "\n---"); found {
+			body = strings.TrimPrefix(rest, "\n")
+		}
+	}
+	if strings.TrimSpace(body) == "" {
+		return ""
+	}
+	sum := sha256.Sum256([]byte(body))
+	return hex.EncodeToString(sum[:])
+}
+
+// storedSourceHash reads source_hash from the source page a previous ingest
+// wrote, or "" when there is no such page or no such field.
+//
+// Absence always means "extract". That is the fail-safe direction: a missed
+// skip costs one extraction, while a wrong skip loses an update silently — and
+// silence is the failure mode this repo keeps paying for.
+func storedSourceHash(brainDir, sourceSlug string) string {
+	raw, err := os.ReadFile(filepath.Join(brainDir, "wiki", "sources", sourceSlug+".md"))
+	if err != nil {
+		return ""
+	}
+	for _, line := range strings.Split(string(raw), "\n") {
+		if key, value, found := strings.Cut(line, ":"); found && strings.TrimSpace(key) == "source_hash" {
+			return strings.TrimSpace(value)
+		}
+	}
+	return ""
+}
+
+// reviseOrCreate returns the bytes to write for one extracted page: merged into
+// whatever is already on disk, or the page as-is when the path is new.
+//
+// This is the compile half of compile-on-ingest. Before it, the write was an
+// unconditional os.WriteFile, so re-ingesting a source replaced its page
+// wholesale and discarded everything accumulated since — including human edits
+// (hyperguild#43: "re-ingest overwrites wholesale").
+//
+// The merge is deliberately asymmetric in the existing page's favour:
+// wiki.Merge keeps the on-disk frontmatter and the on-disk version of any
+// section both sides carry, adds sections only the extraction produced, and
+// unions bullet lists. So a freshly extracted page can add to a note but never
+// silently rewrite what a human put there — which is what makes it safe to run
+// this on every ingest rather than behind a review gate.
+//
+// Known limitation: because the on-disk frontmatter wins, `last_updated` keeps
+// its first-ingest date on a revised page. Left deliberately — the alternative
+// is this function editing frontmatter it does not own, and no lint reads that
+// field today.
+func reviseOrCreate(dest string, page wiki.Page) (string, error) {
+	existing, err := os.ReadFile(dest)
+	switch {
+	case os.IsNotExist(err):
+		return page.Content, nil
+	case err != nil:
+		return "", fmt.Errorf("read existing %s: %w", page.Path, err)
+	}
+	return wiki.Merge(wiki.Page{Path: page.Path, Content: string(existing)}, page).Content, nil
+}
+
 // mergeAll deduplicates pages by path, merging content from later occurrences.
 func mergeAll(pages []wiki.Page) []wiki.Page {
 	order := make([]string, 0, len(pages))

ingestion/internal/pipeline/pipeline_test.go

@@ -171,3 +171,125 @@ func mustJSON(v any) string {
 	}
 	return string(b)
 }
+
+// --- B1: compile-on-ingest — revise the existing page, never clobber it -----
+
+func TestRunRaw_MergesIntoAnExistingPageInsteadOfOverwritingIt(t *testing.T) {
+	// Before this, buildAndWrite did an unconditional os.WriteFile: re-ingesting
+	// a source replaced its page wholesale, discarding anything accumulated
+	// since — including human edits. hyperguild#43 recorded the symptom as
+	// "re-ingest overwrites wholesale".
+	dir := t.TempDir()
+	page := filepath.Join(dir, "wiki", "sources", "shape-up.md")
+	require.NoError(t, os.MkdirAll(filepath.Dir(page), 0o755))
+	require.NoError(t, os.WriteFile(page, []byte(`---
+title: 'Shape Up'
+---
+
+## Summary
+
+A book about shaping product work.
+
+## My Take
+
+Read this before the next planning cycle.
+`), 0o644))
+
+	_, err := RunRaw(dir, "shape-up", []RawPage{{
+		Title: "Shape Up", Type: "source", Subtype: "book", Domain: "product-strategy",
+		Content: "## Summary\n\nA book about shaping product work.\n\n## Key Claims\n\n- Appetite beats estimation.\n",
+	}}, false)
+	require.NoError(t, err)
+
+	got, err := os.ReadFile(page)
+	require.NoError(t, err)
+	s := string(got)
+
+	assert.Contains(t, s, "## My Take", "a section the extraction never saw must survive the merge")
+	assert.Contains(t, s, "Read this before the next planning cycle.")
+	assert.Contains(t, s, "Appetite beats estimation.", "newly extracted material must still land")
+}
+
+func TestRunRaw_ReIngestingUnchangedContentIsByteIdentical(t *testing.T) {
+	// Compile-on-ingest runs on every ingest, so "nothing changed" must produce
+	// no diff at all. Otherwise every re-ingest dirties the tree, brain-sync
+	// commits it, and the git history fills with noise that hides real edits.
+	dir := t.TempDir()
+	raw := []RawPage{{
+		Title: "Shape Up", Type: "source", Subtype: "book", Domain: "product-strategy",
+		Content: "## Summary\n\nA book about shaping product work.\n\n## Open Questions\n\n- Does appetite scale?\n",
+	}}
+
+	_, err := RunRaw(dir, "shape-up", raw, false)
+	require.NoError(t, err)
+	first, err := os.ReadFile(filepath.Join(dir, "wiki", "sources", "shape-up.md"))
+	require.NoError(t, err)
+
+	_, err = RunRaw(dir, "shape-up", raw, false)
+	require.NoError(t, err)
+	second, err := os.ReadFile(filepath.Join(dir, "wiki", "sources", "shape-up.md"))
+	require.NoError(t, err)
+
+	assert.Equal(t, string(first), string(second),
+		"re-ingesting unchanged content must leave the page byte-identical")
+}
+
+// --- #105: source drift — an unchanged re-ingest must not re-extract --------
+
+func TestRun_SkipsExtractionWhenTheSourceIsUnchanged(t *testing.T) {
+	// #103 made a re-ingest harmless (byte-identical merge). It is still not
+	// free: the full LLM extraction runs over every chunk to produce a result
+	// that merges to nothing. This is the cost half.
+	dir := t.TempDir()
+	var calls int
+	cfg := Config{Complete: func(_ context.Context, _, _ string) (string, error) {
+		calls++
+		return `[{"title":"Shape Up","type":"source","subtype":"book","domain":"product","content":"## Summary\n\nA book.\n"}]`, nil
+	}}
+
+	content := "---\nsource: https://example.com/shape-up\n---\n\nThe body of the article.\n"
+
+	first, err := Run(context.Background(), cfg, dir, content, "shape-up", false)
+	require.NoError(t, err)
+	require.Equal(t, 1, calls, "first ingest must extract")
+	assert.False(t, first.Skipped)
+
+	second, err := Run(context.Background(), cfg, dir, content, "shape-up", false)
+	require.NoError(t, err)
+	assert.Equal(t, 1, calls, "unchanged content must not be extracted again")
+	assert.True(t, second.Skipped, "the skip must be reported, not silent")
+}
+
+func TestRun_ReExtractsWhenTheSourceBodyChanged(t *testing.T) {
+	dir := t.TempDir()
+	var calls int
+	cfg := Config{Complete: func(_ context.Context, _, _ string) (string, error) {
+		calls++
+		return `[{"title":"Shape Up","type":"source","subtype":"book","domain":"product","content":"## Summary\n\nA book.\n"}]`, nil
+	}}
+
+	_, err := Run(context.Background(), cfg, dir, "---\nsource: x\n---\n\nversion one\n", "shape-up", false)
+	require.NoError(t, err)
+	_, err = Run(context.Background(), cfg, dir, "---\nsource: x\n---\n\nversion two — the source was edited\n", "shape-up", false)
+	require.NoError(t, err)
+
+	assert.Equal(t, 2, calls, "changed content must be re-extracted")
+}
+
+func TestRun_FrontmatterOnlyChangeIsNotDrift(t *testing.T) {
+	// The hash covers the body, so re-fetching a page whose metadata moved
+	// (a re-crawl stamping a new date) is not treated as the source changing.
+	dir := t.TempDir()
+	var calls int
+	cfg := Config{Complete: func(_ context.Context, _, _ string) (string, error) {
+		calls++
+		return `[{"title":"Shape Up","type":"source","subtype":"book","domain":"product","content":"## Summary\n\nA book.\n"}]`, nil
+	}}
+
+	_, err := Run(context.Background(), cfg, dir, "---\nsource: x\nfetched: monday\n---\n\nsame body\n", "shape-up", false)
+	require.NoError(t, err)
+	_, err = Run(context.Background(), cfg, dir, "---\nsource: x\nfetched: tuesday\n---\n\nsame body\n", "shape-up", false)
+	require.NoError(t, err)
+
+	assert.Equal(t, 1, calls, "only the body counts as the source")
+}

ingestion/internal/wiki/merge.go

@@ -34,7 +34,7 @@ type section struct {
 // all others keep a's version. Sections in b not present in a are appended.
 func Merge(a, b Page) Page {
 	fmA, secsA := parseSections(a.Content)
-	_, secsB := parseSections(b.Content)
+	fmB, secsB := parseSections(b.Content)
 
 	idx := make(map[string]int, len(secsA))
 	for i, s := range secsA {
@@ -53,11 +53,60 @@ func Merge(a, b Page) Page {
 		case bulletSections[sB.heading]:
 			secsA[i].content = mergeBullets(sA.content, sB.content)
 		case appendSections[sB.heading]:
-			secsA[i].content = strings.TrimRight(sA.content, "\n") + "\n\n" + strings.TrimLeft(sB.content, "\n")
+			secsA[i].content = appendNew(sA.content, sB.content)
 		}
 	}
 
-	return Page{Path: a.Path, Content: rebuildContent(fmA, secsA)}
+	return Page{Path: a.Path, Content: rebuildContent(refreshOwned(fmA, fmB), secsA)}
+}
+
+// generatorOwned names the frontmatter fields the ingest pipeline may refresh
+// on a page it is revising. Everything else on disk wins.
+//
+// #103 made the on-disk frontmatter win wholesale, which is what stops a
+// regenerating writer from overwriting human-set metadata — a title someone
+// fixed, or the claim-property envelope (brain#21). The cost was that fields
+// the generator legitimately owns froze at their first-ingest values: when it
+// last ran, and what it last saw.
+//
+// An explicit allowlist rather than a carve-out: the ownership boundary is
+// nameable, and a field nobody has thought about stays human-owned by default,
+// which is the safe direction.
+var generatorOwned = map[string]bool{
+	"last_updated": true,
+	"source_hash":  true,
+}
+
+// refreshOwned returns a's frontmatter with the generator-owned fields taken
+// from b where b supplies them. Fields absent from b keep a's value, so a
+// generator that stops emitting a field cannot blank it.
+func refreshOwned(a, b string) string {
+	bVals := make(map[string]string)
+	for _, line := range strings.Split(b, "\n") {
+		key, value, found := strings.Cut(line, ":")
+		if !found || !generatorOwned[strings.TrimSpace(key)] {
+			continue
+		}
+		bVals[strings.TrimSpace(key)] = strings.TrimSpace(value)
+	}
+	if len(bVals) == 0 {
+		return a
+	}
+
+	out := make([]string, 0, len(strings.Split(a, "\n")))
+	for _, line := range strings.Split(a, "\n") {
+		key, _, found := strings.Cut(line, ":")
+		k := strings.TrimSpace(key)
+		if found && generatorOwned[k] {
+			if v, ok := bVals[k]; ok {
+				out = append(out, k+": "+v)
+				delete(bVals, k)
+				continue
+			}
+		}
+		out = append(out, line)
+	}
+	return strings.Join(out, "\n")
 }
 
 func parseSections(markdown string) (frontmatter string, sections []section) {
@@ -96,15 +145,64 @@ func parseSections(markdown string) (frontmatter string, sections []section) {
 	return
 }
 
+// rebuildContent serialises sections back into a page.
+//
+// Each section body is normalised to exactly one trailing newline and no
+// leading blank line. Without that, a section keeps whatever blank lines it was
+// parsed with and this function adds its own, so every merge pass grows the page
+// by a blank line — invisible in a diff viewer, fatal to byte-stability. Compile-on-ingest
+// merges on every run, so "merging changes nothing when nothing changed" has to
+// hold at the byte level, not just the content level.
 func rebuildContent(frontmatter string, sections []section) string {
 	var sb strings.Builder
 	sb.WriteString(frontmatter)
 	for _, sec := range sections {
-		fmt.Fprintf(&sb, "\n## %s\n\n%s", sec.heading, sec.content)
+		body := strings.Trim(sec.content, "\n")
+		fmt.Fprintf(&sb, "\n## %s\n\n%s\n", sec.heading, body)
 	}
 	return sb.String()
 }
 
+// appendNew concatenates the lines of b that a does not already carry.
+//
+// Append sections used to concatenate wholesale, which is correct exactly once.
+// Compile-on-ingest merges a freshly extracted page into the one already on
+// disk, so an unchanged source re-ingested would grow "Open Questions" and
+// "Updates" on every run until the page is unreadable — and the drift guard
+// that would otherwise catch it (hashing the source body) does not exist yet,
+// so idempotency has to hold here rather than upstream.
+//
+// Line-level rather than block-level: the common case is a re-ingest carrying
+// one genuinely new line alongside several it already has. A block-level
+// "skip if identical" would duplicate the old lines every time that happened.
+//
+// The cost is that a line repeated verbatim on purpose collapses to one. For
+// these sections — questions, updates, evolving notes — a repeat is noise
+// rather than signal, and the failure is conservative: the page stops growing
+// instead of growing wrongly.
+func appendNew(a, b string) string {
+	seen := make(map[string]bool)
+	for _, line := range strings.Split(a, "\n") {
+		if t := strings.TrimSpace(line); t != "" {
+			seen[t] = true
+		}
+	}
+
+	var add []string
+	for _, line := range strings.Split(b, "\n") {
+		t := strings.TrimSpace(line)
+		if t == "" || seen[t] {
+			continue
+		}
+		seen[t] = true
+		add = append(add, line)
+	}
+	if len(add) == 0 {
+		return a // nothing new — leave the section byte-identical
+	}
+	return strings.TrimRight(a, "\n") + "\n\n" + strings.Join(add, "\n") + "\n"
+}
+
 func mergeBullets(a, b string) string {
 	seen := make(map[string]bool)
 	var lines []string

ingestion/internal/wiki/merge_test.go

@@ -53,3 +53,93 @@ func TestMerge_KeepsFrontmatterFromA(t *testing.T) {
 	assert.Contains(t, got.Content, "title: A")
 	assert.NotContains(t, got.Content, "title: B")
 }
+
+// --- brain#21/B1: merging must be safe to run again -------------------------
+
+func TestMerge_AppendSectionDoesNotDuplicateContentAlreadyPresent(t *testing.T) {
+	// Compile-on-ingest merges a freshly extracted page into the one already
+	// on disk. Bullet sections union, so they are naturally idempotent — but
+	// append sections concatenate, so re-ingesting an unchanged source would
+	// grow "Open Questions" on every run until the page is unreadable.
+	page := Page{Path: "wiki/sources/x.md", Content: `---
+title: X
+---
+
+## Open Questions
+
+- Does this converge?
+`}
+
+	once := Merge(page, page)
+	twice := Merge(once, page)
+
+	assert.Equal(t, 1, strings.Count(once.Content, "Does this converge?"),
+		"merging a page with itself must not duplicate an append section")
+	assert.Equal(t, once.Content, twice.Content,
+		"merge must be idempotent — a third pass may not keep growing the page")
+}
+
+func TestMerge_AppendSectionStillAccumulatesGenuinelyNewContent(t *testing.T) {
+	// The dedup must not turn append sections into keep-a-only sections:
+	// new material is exactly what compile-on-ingest exists to add.
+	existing := Page{Path: "wiki/sources/x.md", Content: `---
+title: X
+---
+
+## Open Questions
+
+- Does this converge?
+`}
+	fresh := Page{Path: "wiki/sources/x.md", Content: `---
+title: X
+---
+
+## Open Questions
+
+- What happens under load?
+`}
+
+	got := Merge(existing, fresh).Content
+	assert.Contains(t, got, "Does this converge?", "existing question must survive")
+	assert.Contains(t, got, "What happens under load?", "new question must be added")
+}
+
+func TestMerge_GeneratorOwnedFrontmatterFieldsAreRefreshed(t *testing.T) {
+	// #103 made on-disk frontmatter win wholesale, which is what protects
+	// human-set fields from a regenerating writer. The cost was that fields the
+	// generator legitimately owns — when it last ran, what it last saw — froze
+	// at their first-ingest values. An explicit owned-set is the boundary:
+	// named fields the generator may refresh, everything else human-owned.
+	existing := Page{Path: "wiki/sources/x.md", Content: `---
+title: 'X'
+last_updated: 2026-01-01
+source_hash: aaaa
+half_life: durable
+claim_provenance: asserted
+---
+
+## Summary
+
+Original.
+`}
+	fresh := Page{Path: "wiki/sources/x.md", Content: `---
+title: 'Renamed by the model'
+last_updated: 2026-08-22
+source_hash: bbbb
+---
+
+## Summary
+
+Rewritten by the model.
+`}
+
+	got := Merge(existing, fresh).Content
+
+	assert.Contains(t, got, "last_updated: 2026-08-22", "generator owns last_updated")
+	assert.Contains(t, got, "source_hash: bbbb", "generator owns source_hash")
+	assert.Contains(t, got, "title: 'X'", "a human-set title must survive regeneration")
+	assert.Contains(t, got, "half_life: durable", "the claim envelope must survive regeneration")
+	assert.Contains(t, got, "claim_provenance: asserted")
+	assert.Contains(t, got, "Original.", "an existing section still wins over the model's rewrite")
+	assert.NotContains(t, got, "Renamed by the model")
+}