slate-ar-quality - SKILL.md Agent Skill

description: Slate v2 quality-gap Autoresearch shortcut. Runs deep-research and quality-gap loops for API/DX/architecture/test coverage gaps, then routes accepted findings to slate-patch, slate-plan, slate-ar-gate, or slate-ar-perf. argument-hint: '[slug/topic | quality gap target]' disable-model-invocation: true name: slate-ar-quality metadata: skiller: source: .agents/rules/slate-ar-quality.mdc

Slate AR Quality

Handle $ARGUMENTS by loading slate-ar and running its quality-gap mode.

Use this when the user wants to perfect a Slate v2 surface beyond a single bug: API/DX gaps, architecture gaps, missing behavior coverage, docs/reference holes, example quality, or broad "make this better" research.

Contract

Target cwd is .tmp/slate-v2.
Use generic Codex Autoresearch research and quality-gap mechanics through slate-ar.
Keep durable state under .tmp/slate-v2/autoresearch.research/<slug>/.
quality_gap=0 closes the current accepted checklist only. It does not mean discovery is permanently complete.

Commands

Use Codex Autoresearch directly through the slate-ar wrapper:

<autoresearch-cli> research-setup --cwd .tmp/slate-v2 --slug "<slug>" --goal "<goal>"
<autoresearch-cli> quality-gap --cwd .tmp/slate-v2 --research-slug "<slug>" --list
<autoresearch-cli> gap-candidates --cwd .tmp/slate-v2 --research-slug "<slug>"

Use gap-candidates --apply only after inspecting the candidates and deciding the write scope is safe.

Routing

Correctness gap with missing oracle: slate-patch.
Test/behavior suite gap where the oracle exists: slate-ar-gate.
Performance gap: slate-ar-perf.
Public API/runtime architecture gap: slate-plan.
Generic measured loop gap: slate-ar.

Do not use this as a substitute for a direct bug fix. If the failing behavior is already known and reproducible, use slate-patch.