菜单
Todo-first progress convention for multi-step procedures — loaded by every multi-step agent to track its own steps without drift
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
Keep a Changelog methodology — loaded by the ship phase to update CHANGELOG.md with user-facing changes, filtering out internal-only commits
Generator-evaluator separation and review methodology — loaded by review agents to enforce fresh-context review discipline, Conventional Comments format, and gate verdicts
Architecture Decision Record creation and management — loaded by planner and orchestrator when significant technical decisions need to be recorded with context and consequences
Adversarially review a technical design document with fresh context before the human gate. Dispatches the built-in `general-purpose` subagent (clean context, no shared history with the design-author) against `docs/plans/<id>/design.md` and presents its verdict — APPROVE, REQUEST CHANGES, or COMMENT. Optional, not part of the QRSPI pipeline. Trigger on "review the design doc", "audit design.md", "is this design ready", or `/eng-design-doc-review`.
Git commit discipline methodology — loaded by the ship phase to produce well-formed commits following conventional commits format, the 50/72 rule, and atomic commit principles
Optional PRD methodology — loaded by the questioner agent when a feature request is vague or complex enough to warrant a structured product spec alongside task.md. Produces a PRD artifact that downstream design-author work can ground decisions in.
| name | progress-tracking |
| description | Todo-first progress convention for multi-step procedures — loaded by every multi-step agent to track its own steps without drift |
| user-invocable | false |
A convention, not a gate. It produces no artifact and blocks nothing. It shapes how any agent or skill executing a multi-step procedure keeps its own work visible, so steps are not silently skipped.
When a procedure has two or more ordered steps, seed one todo item per step before starting and mark each complete as you go. A single-step procedure is exempt — a one-item ledger is noise.
One todo item per numbered step — not per phase, not per file. Mark each item
in_progress when its step starts and completed when that step lands,
matching how team-fix already marks each step. Keep items at the
granularity of the procedure's own numbered list.
The orchestrator (the main Claude Code session) owns the single phase-level ledger. An agent executing a multi-step skill tracks its own sub-steps within its own context and never merges them up into the orchestrator's phase ledger. A standalone, directly-invoked skill seeds its own ledger. The two ledgers live in separate ownership scopes and are never read across.
The orchestrator's phase-level TodoWrite contract lives in
skills/qrspi-workflow/SKILL.md — the phase-ledger sibling of this
per-procedure convention.
