Adds the /evaluate-library skill runbook at .claude/skills/evaluate-library/
with SKILL.md (8-filter + 3-prompt protocol, collect-cheap-skip-expensive
ordering, trace-write step, skip sentinel), POLICY.md (ADR-022 summary
≤2 pages), TRACE-TEMPLATE.md (complete YAML frontmatter + 11 headings in
order), and EXAMPLES/ with one approved (clsx) and one rejected
(trpc-to-openapi, named-consumer: fail) worked trace.
Updates session-start.sh to surface the skill in session pointers.
The skill is auto-registered by the harness on SKILL.md creation.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Registers .claude/hooks/library-policy-nudge.sh under PreToolUse/Bash
and PostToolUse/Edit|Write|MultiEdit. The hook emits a non-blocking
system-reminder pointing at /evaluate-library before runtime deps are
added via pnpm add or via direct package.json edits, so policy
evaluation happens before the pre-commit gate fires.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Adapts mattpocock/skills/engineering/improve-codebase-architecture to
this repo. Four files at .claude/skills/improve-codebase-architecture/:
SKILL.md (104 lines):
- Explore -> Present candidates -> Grilling loop process
- "Hard constraints (do not propose violations)" section
enumerating ADRs 006/008/010/012/013/014/015/017/020/021 that
bound the design space
- Repointed at docs/glossary.md (not CONTEXT.md) and
docs/decisions/ (not docs/adr/)
- Exploration shortcuts specific to this repo: pnpm fallow,
pnpm coverage:diff, feature.manifest.ts, pnpm turbo boundaries
- Grilling loop side-effects target the right glossary section
and the next available ADR number (currently 022)
DEEPENING.md (93 lines):
- 4 dependency categories mapped to this repo's reality:
Cat 1 (in-process) -> entities/use-cases/presenters
Cat 2 (local-substitutable) -> our existing real + mock
adapter pattern (every port has both; mocks ARE stand-ins)
Cat 3 (remote but owned) -> cross-feature events via
IEventBus (E0/E1 rules)
Cat 4 (true external) -> Payload, Sentry/OTel, socket.io
(each constrained to its vendor-isolation seam by ADR)
- Seam discipline section recognises DI symbols + manifest entries
as concrete seams alongside .interface.ts files
- Testing strategy: replace not layer (matches ADR-020 L0 + L1)
- Conformance check command list at the end (typecheck, lint,
test --coverage, conformance, fallow:audit, coverage:diff)
INTERFACE-DESIGN.md (66 lines):
- Parallel sub-agent "Design It Twice" pattern preserved
- Every sub-agent brief MUST include glossary terms + ADR
constraints + manifest awareness
- Output items extended with "Manifest + binder impact" and
"ADR conflicts (if any)"
- Comparison axes include conformance impact + coverage delta
- Cross-feature moves flag release-please version-bump
implications (per ADR-021 commit-path targeting)
LANGUAGE.md (79 lines):
- Matt's 7 abstract terms preserved (module, interface,
implementation, depth, seam, adapter, leverage, locality)
- New "Mapping to this repo's identifiers" table — abstract
term -> concrete file shape (e.g. seam -> *.interface.ts +
DI symbol + manifest entry + <gen:*> anchor)
- Rejected framings extended with our reserved meanings
("boundary" stays the ESLint workspace-tag term; "service"
stays the DI port term)
Per user follow-up: vocabulary anchored so that "module" defaults
to "feature" in this repo (since features are our primary unit of
organisation). Abstract refactor sense survives only when the cross-
scale abstraction is the point. Glossary.md updated:
- "Feature" entry adds the "module = feature in refactor sense"
cross-link
- New "Architecture refactor vocabulary" section with 9 terms
(Module, Interface (refactor sense), Implementation, Depth,
Seam, Adapter, Leverage, Locality, Deletion test, Deepening)
— all framed so feature is the primary instance
- Flagged ambiguities entry for "module" rewritten to capture the
three coexisting senses (workspace package / Node ESM / refactor
vocabulary defaulting to feature); new entries for "seam" and
"adapter" to prevent drift with the existing "boundary" / "service"
/ "scope" reservations
Hooks updated:
- session-start.sh skills line lists the new skill
- prompt-context.sh adds a 10th keyword group firing on
refactor / deepening / shallow / architecture / seam / adapter /
interface design / design it twice — inject points at SKILL.md
+ summarises the vocabulary and hard constraints
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Closes the user's ask: versioning + a changelog generated on merging
to main, building on the just-mandated Conventional Commits substrate
(CLAUDE.md Key Conventions).
Architecture: ADR-021. Cookbook: docs/guides/releasing.md.
Initial state — six tracked packages at v0.1.0:
- . -> template-vertical (tag: template-v...)
- packages/auth -> @repo/auth (tag: auth-v...)
- packages/blog -> @repo/blog (tag: blog-v...)
- packages/media -> @repo/media (tag: media-v...)
- packages/marketing-pages -> @repo/marketing-pages (tag: marketing-pages-v...)
- packages/navigation -> @repo/navigation (tag: navigation-v...)
Core packages, tooling, and apps are NOT independently versioned
(ADR-021 rationale: core bumps cascade; apps aren't consumables;
surfacing them would create noise without information).
Configuration:
- release-please-config.json - 6 tracked packages, hybrid scope,
pre-1.0 conservative bump policy
(feat: -> patch, feat!: -> minor),
conventional-commit type mapping
- .release-please-manifest.json - baseline 0.1.0 for all 6 packages
- .github/workflows/release-please.yml - googleapis/release-please-
action@v4 on push to main,
concurrency-gated, write
permissions for the rolling PR
Workflow: on every push to main, release-please scans commits since
the last release tag PER PACKAGE (using commit-path, not the
conventional-commit scope), updates a single rolling release PR with
version bumps + per-package CHANGELOG entries. Merging that PR cuts
per-package tags + GitHub releases.
CHANGELOG files seeded at v0.1.0 baseline:
- CHANGELOG.md (root)
- packages/<feature>/CHANGELOG.md (5 features)
Subsequent versions are appended by release-please from commit
history. Do not edit manually.
Visibility surfaces updated (every agent entry point):
- CLAUDE.md Read First + new "Versioning is hybrid" Key Conventions
bullet (with bump policy summary)
- AGENTS.md preamble - new "Releases:" callout alongside Commits
- docs/glossary.md - new Releasing section with 8 terms (Conventional
Commits, release-please, Hybrid versioning, Tag prefix, Rolling
release PR, Bump targeting, Pre-1.0 bump policy, Release-As trailer,
CHANGELOG.md)
- docs/README.md - guides tree updated with releasing.md
- .claude/hooks/session-start.sh - one-line release reminder
- .claude/hooks/prompt-context.sh - new keyword group for
release/version/bump/semver/tag prompts
Package.json version bumps:
- root: name "template" -> "template-vertical", version "0.1.0"
- packages/auth, blog, media, marketing-pages, navigation: "0.0.0" -> "0.1.0"
Root rename rationale: release-please tags use the package-name + the
component prefix; "template-vertical" matches the repo identity (and
the user's question preview).
First release-please PR after this lands will sweep all subsequent
post-baseline commits into 0.1.1 / 0.2.0 bumps as appropriate.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The implementer + reviewer sandcastle prompts already enforce
conventional commits, but the convention was buried in those prompt
templates — agents not running through sandcastle (e.g. interactive
Claude sessions, ad-hoc fixups) had no visible signal that the format
is non-negotiable.
Four visibility surfaces now carry the signal:
- CLAUDE.md Key Conventions: new top bullet stating the spec
(<type>(<scope>): <subject>), the full type list, breaking-change
syntax (!), and three example commits
- AGENTS.md preamble: parallel callout alongside the vocabulary
and quality-gates notes
- .claude/hooks/session-start.sh: one-line reminder in the boot
pointers (every session now sees it at start)
- .claude/hooks/prompt-context.sh: new 10th keyword group fires
when a user prompt mentions commit/message/changelog/conventional,
injecting the full spec into the turn's context
Conventional Commits spec: https://www.conventionalcommits.org/
Verified: the prompt-context hook smoke-tests green when a prompt
containing "commit message" is piped in — it emits the conventional
commits pointer.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Architecture record for the agent-first coverage initiative seeded by
the 2026-05-13 PRD. Captures the durable decisions:
- 4-layer architecture (L0 vitest, L1 diff, L2 aggregate, L3 mutation)
- Manifest-driven coverage band as single source of truth (vitest +
assertFeatureConformance + pnpm coverage:diff all read from it)
- Cover-the-diff (changed lines), not cover-the-new-code
- Committed coverage/summary.json (no SaaS), trend via git log
- Mutation testing scoped to entities + use-cases, on-demand only
- Machine-first output format (JSON stdout, human stderr)
Glossary gets a new "Coverage" section with 7 entries (coverage band,
L0-L3 layers, diff coverage, mutation testing, mutation score,
coverage/summary.json), plus two relationship rows and a flagged
ambiguity for "coverage" qualifiers.
prompt-context.sh hook gets a 9th keyword group — when a prompt
mentions coverage / uncovered / lcov / mutation / stryker, the
relevant ADR + guide path are injected as additional context for
the turn.
This is the documentation layer of the coverage epic. Implementation
(manifest schema, vitest auto-derive, scripts, boot assertion,
mutation tooling) lands in subsequent stories.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Project-level Claude Code hooks committed to .claude/settings.json with
scripts under .claude/hooks/. Three tiers:
Tier 1 — hard guards (exit 2 to block the tool call):
- bash-guard.sh: blocks bypass flags (verify-skip, sign-skip), forceful
push variants, destructive resets, force clean, working-tree-wipe
checkouts/restores, force branch delete, amend, and rm -rf against
root or home. Reinforces CLAUDE.md Git Safety Protocol.
- generator-first-nudge.sh: blocks creating a new top-level
packages/<name> or apps/<name> directory by hand. Allows working
inside an existing package. Reinforces the non-negotiable
generator-first rule.
Tier 2 — context injection (stdout becomes additional context):
- session-start.sh: prints glossary, AGENTS.md, workflow CLI, and
conformance pointers on session boot.
- prompt-context.sh: keyword-matches the user prompt against eight
concept groups (events, realtime, audit, instrumentation, manifest,
workflow, DI, boundaries) and injects the relevant ADR + rule
pointers for the turn.
Tier 3 — side-effect automation:
- post-manifest-edit.sh: when Edit/Write touches feature.manifest.ts,
prints the manifest-first ordering reminder plus the per-feature
verify commands.
- stop-check-manifest-tests.sh: at agent Stop time, if the working
tree has manifest changes but no sibling test changes, exits 2 to
force continuation. Loop-guarded via stop_hook_active.
All hooks are bash + jq, use CLAUDE_PROJECT_DIR for safety, and were
smoke-tested end-to-end (block + allow paths both verified).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
