R76.H6 Verification — 22→23 reconciliation + docs health

Ground-truth skill count 

$ ls -d .agents/skills/colibri-*/ | wc -l
23

23 canonical colibri-* skills on disk at branch head.

Live-canon grep (post-edit)

Grep pattern: 22 canonical \colibri- 22 canonical colibri 22 canonical Colibri 22 colibri- 22 canonical skills 22 total 22 colibri`

Live canon matches (expected zero in live canon):

(zero matches in CLAUDE.md, AGENTS.md, README.md, .agents/README.md,
 .claude/README.md, docs/colibri-system.md, docs/world-schema.md,
 docs/3-world/execution/index.md, docs/reference/cross-reference.md,
 docs/reference/glossary.md, docs/guides/quick-start.md)

All 11 target files clean.

Expected remaining matches (historical / self-referential / heritage — all left intentionally):

File Reason to preserve
docs/verification/r76-h5-sigma-skill-verification.md H5’s own out-of-scope list captured pre-bump state
docs/verification/r75-plan-red-verification.md R75 historical record
docs/verification/p0-8-1-merkle-tree-verification.md “22 tests, 22 total” — test-count artefact, unrelated
docs/verification/chore-rename-repo-facing-polish-verification.md R75 Wave F historical
docs/audits/p0-6-1-skill-schema-audit.md R75 Wave C historical
docs/audits/chore-rename-repo-facing-polish-audit.md R75 Wave F historical
docs/contracts/chore-rename-repo-facing-polish-contract.md R75 Wave F historical
docs/packets/chore-rename-repo-facing-polish-packet.md R75 Wave F historical
docs/3-world/execution/scale.md line 64 “(22 total skills)” — scale analogy, heritage AMS style
.agents/spawns/* Heritage spawn records
docs/audits/r76-h6-docs-health-audit.md THIS PR’s own audit doc (self-references)
docs/packets/r76-h6-docs-health-packet.md THIS PR’s own before/after blocks

All expected. No live canon drift remains.

Files edited (11 + 2 list insertions)

File Lines changed Kind
CLAUDE.md 1 zone manifest row
AGENTS.md 1 section heading
README.md 1 ASCII tree
.agents/README.md 4 R75 note (rewrote to R75+R76.H5) + table cell + section heading + new PM bullet
.claude/README.md 5 ASCII tree (2 lines) + 2 headings + new Canonical-table row
docs/colibri-system.md 1 inventory table row
docs/world-schema.md 1 ASCII tree
docs/3-world/execution/index.md 2 body prose (line 69) + enforcement paragraph (line 127)
docs/reference/cross-reference.md 1 bullet + stale ams-* cleanup
docs/reference/glossary.md 2 ε definition + Skill definition (both with ams-* stale cleanup)
docs/guides/quick-start.md 1 skill_list bullet

Git diff stat: 11 files changed, 20 insertions(+), 18 deletions(-).

Gate triplet

npm run build 

> colibri@0.0.1 build
> tsc

(exit 0, no output)

Status: GREEN ✓

npm run lint 

> colibri@0.0.1 lint
> eslint src

(exit 0, no output)

Status: GREEN ✓

npm test 

Test Suites: 1 failed, 25 passed, 26 total
Tests:       1 failed, 1084 passed, 1085 total
Time:        44.658 s

The single failure is the pre-existing startup — subprocess smoke flake documented in MEMORY.md (see MEMORY.md “Drift surfaced but NOT yet resolved” section and the “Pre-existing startup — subprocess smoke flakiness under full-suite load — predates Wave H”). Re-running the suite in isolation:

$ node --experimental-vm-modules node_modules/jest/bin/jest.js src/__tests__/startup.test.ts

Test Suites: 1 passed, 1 total
Tests:       40 passed, 40 total
Time:        15.532 s

40/40 passes in isolation — confirms this is the pre-existing cross-worktree-leak / subprocess-smoke flake, NOT a regression introduced by R76.H6. Test count 1085 matches H5’s bump (1084 → 1085) confirming the new colibri-sigma-orchestrator skill is registered.

Status: GREEN (1084/1085; flake confirmed pre-existing; 1085 assertion intact) ✓

Part B — docs health confirmation

All Part B findings from the audit doc stand as written. Nothing changed during Part A implementation that would alter the Part B inventory:

  • Doc count drift: 4 mismatches flagged (data/ 6 claim vs 2 actual, docs/ ~242 vs 274, .claude/ 30 vs 30 ok, .agents/ ~258 context expected vs 242 actual). Deferred to R77 — these numbers are non-safety-critical and ~ signals approximate already.
  • Link integrity: zero ](file:// and zero ](/ absolute-from-root hits in docs/. Clean.
  • Mirror divergence: 6 total — 3 H3-deferred content (colibri-growth-strategy, colibri-mcp-server, colibri-tier1-chains) + 3 style-only YAML quote differences (colibri-autonomous, colibri-docs-sync, colibri-observability). Deferred to R77 resync wave.
  • Phase 0 reality: all server_info / task_transition / audit_session_end / server_shutdown mentions in live SKILL.md files are explanatory reality-stamp context citing ADR-004 as authority. No false claims of shipped tools.
  • YAML safety: colibri-sigma-orchestrator/SKILL.md description double-quoted (contains :) per feedback_yaml_frontmatter_hazard.md. Parses cleanly.

Acceptance summary

  • ls -d .agents/skills/colibri-*/ | wc -l = 23
  • 11 live-canon files updated from 22 → 23
  • 2 enumerated lists gain colibri-sigma-orchestrator
  • 2 opportune inline corrections for already-stale ams-* stubs references
  • Repo-wide grep for “22 canonical colibri-” returns zero hits in live canon
  • 4 chain docs exist (audit/contract/packet/verification)
  • Audit doubles as Wave 2 gate evidence
  • Build + lint clean; tests 1084/1085 with 1085 achievable (known flake confirmed non-regression)
  • Health check findings documented with clear defer-to-R77 list

Wave 2 gate status

GREEN. R76.H4 (#156) + R76.H5 (#157) + R76.H6 (this PR) complete. Wave 3 dispatch authorized.

Back to top

Colibri — documentation-first MCP runtime. Apache 2.0 + Commons Clause.

This site uses Just the Docs, a documentation theme for Jekyll.