Documentation Accuracy & Skill Refinement
Fixes several inaccuracies in the README that predated the v1.2.1 system card hardening work, plus refines the new linting-instructions skill scope.
README Fixes
Agent Model Table (was incorrect since v1.0)
The agent table incorrectly listed engineer agents as opus. The actual 3-tier model routing is:
| Tier | Model | Agents | Purpose |
|---|---|---|---|
| Deep analysis | opus | go-qa, go-impl, py-qa, py-impl, ts-qa, ts-impl
| Security review, implementation correctness (effort: high) |
| Balanced work | sonnet | All engineers, review agents (*-idioms, *-tests, *-simplify), infra, spec-planner, playwright, perplexity, pdf-parser, docs-keeper | Implementation proposals, pattern review, tooling |
| Fast/cheap | haiku | go-docs, py-docs, ts-docs, web-docs
| Documentation review |
Skills Table
Added to user-invocable (were missing entirely):
analyzing-usage— Claude Code cost/efficiency analysislinting-instructions— plugin prompt review against Anthropic model cardsusing-git-worktrees— isolated parallel development
Reclassified (were listed as auto-activated but have user-invocable: true):
learning-patterns— moved to user-invocable table
Added to auto-activated (was missing):
playwright-skill— runtime library for testing-e2e
Added to agents table (was missing):
pdf-parser— PDF data extraction agent
Count fix: dev-tools plugin skills 13 → 14
Skill Refinement
linting-instructionsenforcer triggers narrowed to skill/agent authoring context only — won't fire on general coding prompts- Description updated to reference "Anthropic model cards" instead of "system card-derived rules"