chore(work): decompose 2026-05-14-library-evaluation-policy epic

Decomposer produced 9 stories under docs/work/2026-05-14-library-
evaluation-policy/, ordered to land the schema foundation first and
the cross-referencing content (skill, guide, backfill) last:

01 - trace schema module + docs/library-decisions/ foundation
02 - pre-commit check script
03 - Claude PreToolUse / PostToolUse hooks
04 - evaluate-library skill
05 - human guide (docs/guides/adding-a-library.md)
06 - sandcastle reviewer prompt update
07 - generator pre-shipped traces for optional cores
08 - backfill traces for existing runtime deps
09 - CLAUDE.md Key Conventions bullet

Anchored by ADR-022 + the approved PRD at
docs/work/prds/2026-05-14-library-evaluation-policy.prd.md.

docs/work/2026-05-14-library-evaluation-policy/05-human-guide/_story.md

@@ -0,0 +1,37 @@
+---
+id: 05-human-guide
+epic: 2026-05-14-library-evaluation-policy
+title: Human reading-room guide — docs/guides/adding-a-library.md
+type: technical-story
+status: todo
+feature: docs
+depends-on: [04-evaluate-library-skill]
+blocks: [09-claude-md-update]
+---
+
+## Goal
+
+Write `docs/guides/adding-a-library.md` — the human-readable guide explaining the library evaluation policy with worked examples (one approved, one rejected), the tier trigger, the four enforcement layers, and how to invoke the skill. The guide targets a maintainer reading the repo for the first time.
+
+## Why
+
+ADR-022 is the source of truth but is written for decision-record density, not onboarding. The guide translates the policy into a narrative that answers "why does this exist?" and "what do I actually do?" before a developer encounters the pre-commit gate for the first time. Worked examples anchor the abstract filter list to concrete outcomes.
+
+## Done when
+
+- `docs/guides/adding-a-library.md` exists with: (1) a "Why this exists" section explaining the uncodified-surface problem and the three signals from the PRD; (2) the tier trigger (feature/core packages require traces; app-tier does not; devdeps exempt); (3) the four enforcement layers (Claude hook → skill → pre-commit → sandcastle) in latency order; (4) step-by-step "how to add a library" walkthrough pointing at the `/evaluate-library` skill; (5) a worked approved example (brief — the full trace lives in `EXAMPLES/` from Story 04); (6) a worked rejected example (`trpc-to-openapi`, `named-consumer: fail`); (7) a link to ADR-022 and `docs/library-decisions/_template.md`.
+- `pnpm lint && pnpm fallow:audit` pass.
+
+## In scope
+
+- `docs/guides/adding-a-library.md` — the guide document.
+
+## Out of scope
+
+- The skill itself (Story 04) — already landed.
+- CLAUDE.md update (Story 09) — that bullet points here and to ADR-022 but lands separately.
+- Changing any existing guide or ADR.
+
+## Tasks
+
+- [ ] Write `docs/guides/adding-a-library.md` with all seven sections from Done when (why, tier trigger, four layers, how-to walkthrough, worked approved + rejected examples, cross-links); all gates pass on this single commit.