chore(claude): add the engineering toolkit with tiered model pins#5
Merged
Conversation
## Summary Add a `.claude/` toolkit and a top-level `CLAUDE.md` so the autonomous and semi-autonomous workflows (`/lfg` and friends) run against docs-kit's own domain — the `DocsUI::` Phlex kit, the per-site config surface, the registry, the install generator, and the CSS/deploy contracts — instead of another project's internals. Every command and agent pins a model **tier alias** in frontmatter rather than a full model ID, so pins track the latest model per tier and never go stale: - `haiku` — mechanical/config work - `sonnet` — layer specialists / pattern-following implementation (default) - `opus` — orchestration, security, PR review - `fable` — read-only planning that hands execution to cheaper models (`/plan`) ## What's included - Commands: `/plan` (fable), `/lfg` `/architect` `/security` `/review-pr` `/github-review-pr` (opus), `/tdd` `/github-review-comments` `/github-review-failures` (sonnet) - Rules: coding-style, git-workflow, testing, agents - `CLAUDE.md` — critical rules, architecture map, slash-command table, and the model-tier convention - `.claude/README.md` — toolkit index + the tier convention - `.claude/settings.local.json` — permission allowlist (bun/bunx/gh/git/bundle/rake) ## Notes - The perf command/rule was intentionally dropped — docs-kit has no benchmark harness. - All content is rewritten for docs-kit invariants: compose from `DocsUI::` components (never raw daisyUI), read `DocsKit.configuration` (with defaults for backwards compat), progressive enhancement (works with JS off), one `docs-nav` Stimulus controller, `config.themes` in sync with the Tailwind `@plugin` list, and required setup wired into the install generator + `docs-kit new` template. ## Verification - [x] All 9 commands carry a tier alias; agents rule documents cheaper subagent pins - [x] No stray references to other projects' internals - [x] `.claude/` is tracked (not gitignored)
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Adds a
.claude/engineering toolkit and a top-levelCLAUDE.mdso autonomousand semi-autonomous workflows (
/lfgand its specialists) run against docs-kit'sown domain — the
DocsUI::Phlex kit, the per-site config surface, the registry,the install generator, and the CSS/deploy contracts.
Every command and agent pins a model tier alias in frontmatter rather than a
full model ID, so pins track the latest model per tier and never go stale.
fable/plandocs/plans/markdown for cheaper models to executeopus/lfg,/architect,/security,/review-pr,/github-review-prsonnet/tdd,/github-review-comments,/github-review-failureshaikurules/agents.md)Why
claude-opus-4-8) goes stale on every model release;tier aliases (
opus/sonnet/haiku/fable) fix that class of stalenesspermanently by tracking the latest model per tier.
and the review runbooks follow prescriptive prompts and run well a tier down.
/planputs the capable model where it pays off: it audits and writes aself-contained spec (context, ordered steps, verification gates, out-of-scope)
that
/lfgexecutes on a cheaper model without any of the planning session'scontext.
What's included
.claude/commands/):plan,lfg,architect,security,review-pr,github-review-pr,github-review-failures,github-review-comments,tdd..claude/rules/):coding-style,git-workflow,testing,agents.CLAUDE.md: critical rules, architecture map, slash-command table, and themodel-tier convention.
.claude/README.md: toolkit index + tier convention..claude/settings.local.json: permission allowlist.Notes
DocsUI::components(never raw daisyUI), read
DocsKit.configurationwith defaults (backwardscompat), progressive enhancement (works with JS off), one
docs-navStimuluscontroller,
config.themesin sync with the Tailwind@pluginlist, andrequired setup wired into the install generator +
docs-kit newtemplate.harness.
Test plan
.claude/is tracked (not gitignored)Docs/config only — no application code, no specs affected.