1.8 KiB
1.8 KiB
id, prd, title, type, status, features, created
| id | prd | title | type | status | features | created | |||
|---|---|---|---|---|---|---|---|---|---|
| 2026-05-14-library-evaluation-policy | docs/work/prds/2026-05-14-library-evaluation-policy.prd.md | Library evaluation policy — skill, traces, enforcement stack | epic | done |
|
2026-05-14 |
Goal
Implement a four-layer enforcement stack — Claude hook, skill, pre-commit hook, sandcastle reviewer prompt — that makes every new runtime dependency in a feature- or core-tier package produce a permanent library trace at docs/library-decisions/<YYYY-MM-DD>-<package-name>.md. Rejection traces are first-class records. Codifies ADR-022.
Why
The repo's narrow third-party surface is uncodified. New dependencies enter via pnpm add with no checkpoint. Three signals exposed the gap: a near-miss adding a build-time-only library, post-hoc ADR records (002/014/017), and a silent EU-data-residency risk from US-only SaaS defaults. The enforcement stack mirrors the 5-gate conformance pattern — same vocabulary, same agent feedback loop.
Stories
- 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