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.

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/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/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

ingestion/internal/mcp/tools_notes.go

@@ -6,8 +6,11 @@ import (
 	"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"
 )
 
@@ -52,6 +55,20 @@ func noteToolDescriptors() []map[string]any {
 	}
 }
 
+// --- 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"`
 }
@@ -75,8 +92,17 @@ type brainEditArgs struct {
 }
 
 // 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) {
-	return s.editWithScope(args, notes.ScopeNotes)
+	_, out, err := s.editWithScope(args, notes.ScopeNotes)
+	return out, err
 }
 
 // brainEditRoot edits a governance file (CLAUDE.md). De-surfaced AND gated on
@@ -89,22 +115,72 @@ func (s *Server) brainEditRoot(ctx context.Context, args json.RawMessage) (json.
 		return nil, fmt.Errorf("brain_edit_root requires the local static principal; " +
 			"a governance file is not editable from a model-facing origin")
 	}
-	out, err := s.editWithScope(args, notes.ScopeRoot)
-	if err == nil {
-		// A durable instruction rewrite should be loud, not silent.
-		slog.Warn("brain_edit_root applied", "principal", info.principal, "result", string(out))
+	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
 }
 
-func (s *Server) editWithScope(args json.RawMessage, scope notes.Scope) (json.RawMessage, error) {
+// 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 nil, fmt.Errorf("parse args: %w", err)
+		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 nil, err
+		return notes.Result{}, nil, err
 	}
 
 	// Reindex only when the edit could have changed what the index reads.
@@ -121,7 +197,8 @@ func (s *Server) editWithScope(args json.RawMessage, scope notes.Scope) (json.Ra
 			res.ReindexError = ierr.Error()
 		}
 	}
-	return json.Marshal(res)
+	out, merr := json.Marshal(res)
+	return res, out, merr
 }
 
 // wingOf returns the wing for a wiki path, or "" for knowledge/ and root.

ingestion/internal/mcp/tools_notes_test.go

@@ -2,13 +2,22 @@
 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"
@@ -125,3 +134,164 @@ func mustJSON(t *testing.T, v any) []byte {
 	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

@@ -36,11 +36,15 @@ var pathLocks sync.Map // string -> *sync.Mutex
 
 // Result is what a successful edit reports.
 type Result struct {
-	Path         string `json:"path"`
-	ContentHash  string `json:"content_hash"`
-	BytesBefore  int    `json:"bytes_before"`
-	BytesAfter   int    `json:"bytes_after"`
-	ReindexError string `json:"reindex_error,omitempty"`
+	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.
@@ -149,11 +153,13 @@ func Edit(brainDir, path, oldStr, newStr string, scope Scope) (Result, error) {
 	}
 
 	sum := sha256.Sum256([]byte(after))
+	prev := sha256.Sum256(before)
 	return Result{
-		Path:        t.Rel,
-		ContentHash: hex.EncodeToString(sum[:]),
-		BytesBefore: len(before),
-		BytesAfter:  len(after),
+		Path:            t.Rel,
+		PrevContentHash: hex.EncodeToString(prev[:]),
+		ContentHash:     hex.EncodeToString(sum[:]),
+		BytesBefore:     len(before),
+		BytesAfter:      len(after),
 	}, nil
 }
 

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")
+}