菜单
Run the full docs-alignment backfill audit across docs/ + archived REQUIREMENTS.md. Top-level slash command — delegates to reposix-quality-doc-alignment backfill playbook. Dispatches ~25–35 Haiku extractor subagents in waves of 8, runs merge-shards, generates PUNCH-LIST.md.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Use as a reposix orchestrator/coordinator facing one of five concrete situations, unsure whether to act, wait, or escalate — (1) a dispatched child hasn't committed in a while, stuck vs slow?; (2) about to dispatch a fix for a BLOCKER/HIGH whose only evidence is a code reading, not something executed; (3) an intake/ledger row has a fix design already sketched, about to implement as written; (4) the next roadmap item no longer seems right given new information; (5) found out-of-charter work (bug, missing tool, doc lie) and unsure whether to fix now, file, or ask. Also for the fable-consult template or the CONSULT-DECISIONS.md ledger format. NOT for routine dispatch (use coordinator-dispatch) or anything a hook already resolves.
Dispatch template for a reposix phase/debt-drain coordinator — charter to paste, canonical role→subagent_type mapping, model-tier table, lane-slicing checklist, report contract, relief trigger, pause/resume brief. Pull this before dispatching a wave of sub-agents.
Run subjective rubric checks (cold-reader hero clarity, install positioning, headline-numbers sanity) by dispatching one unbiased subagent per stale catalog row in parallel. Reads quality/catalogs/subjective-rubrics.json, persists JSON artifacts to quality/reports/verifications/subjective/<rubric-id>.json. The next runner sweep re-grades the catalog rows. Pre-approved for v0.12.0 P61 per .planning/research/v0.12.0/open-questions-and-deferrals.md line 124.
Spawn the reposix dark-factory regression test — a fresh subprocess agent given only `reposix init` and a goal must complete clone+grep+edit+commit+push using pure git/POSIX, including the conflict-rebase and blob-limit recovery cycles. The StrongDM/dark-factory regression harness for the v0.9.0 architecture. Invoked from CI release-gate and from local dev (/reposix-agent-flow).
Run the docs-alignment dimension — refresh stale rows for a single doc OR run the full backfill audit across docs/ + archived REQUIREMENTS.md. Subagents propose claim→test bindings with file:line citations; the `reposix-quality` binary validates and mints catalog state. Reads quality/catalogs/doc-alignment.json. Top-level only (depth-2 subagent fan-out unreachable from inside /gsd-execute-phase; subscription users cannot fall back to claude -p).
Refresh stale docs-alignment catalog rows for a single doc that drifted. Top-level slash command — delegates to reposix-quality-doc-alignment refresh playbook. Invoke after pre-push BLOCKs with STALE_DOCS_DRIFT.
| name | reposix-quality-backfill |
| description | Run the full docs-alignment backfill audit across docs/ + archived REQUIREMENTS.md. Top-level slash command — delegates to reposix-quality-doc-alignment backfill playbook. Dispatches ~25–35 Haiku extractor subagents in waves of 8, runs merge-shards, generates PUNCH-LIST.md. |
| argument-hint | |
| allowed-tools | ["Bash","Read","Edit","Write","Task","Grep","Glob","Skill"] |
This skill exists as a separate slash command so the user can type /reposix-quality-backfill directly. The full implementation lives in .claude/skills/reposix-quality-doc-alignment/backfill.md.
<cross_references>
.claude/skills/reposix-quality-doc-alignment/SKILL.md — umbrella skill (parses refresh vs backfill)..claude/skills/reposix-quality-doc-alignment/backfill.md — backfill playbook (normative; consumes 06-p65-backfill-brief.md)..planning/research/v0.12.0-docs-alignment-design/06-p65-backfill-brief.md — backfill protocol source of truth..planning/research/v0.12.0-docs-alignment-design/03-execution-modes.md — why this is top-level only.
</cross_references>