菜单
Use when a PRD or spec exists and needs to become an executable backlog. Breaks the document into GitHub issues structured as vertical-slice tracer bullets, marks each as HITL or AFK, and files a final QA issue. Requires the gh CLI.
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 when a question has a factual or best-practice answer but a single response might hallucinate, or when the user asks to "get consensus" or "sanity check this with multiple agents". Spawns N subagents on the same question with varied personas and tallies answer frequency.
Use when a question has real trade-offs and no single right answer: architecture choices, library selection, design direction. Runs a multi-round debate where agents see each other's positions across rounds, converge, and a synthesis is produced. For factual questions, use consensus instead.
Use when triaging an inbox: fetching unprocessed emails, classifying each into a label taxonomy, and applying labels via the provider's API. Provider-agnostic skeleton; scripts/label.py must be wired to your email provider before first use.
Use when a workflow built on an MCP server has stabilized and should become a production skill, or when the user says "convert this MCP workflow to a skill". Produces a skill with a bundled script that calls the underlying API directly, eliminating the per-session MCP token overhead.
Use when implementing any task that should end in a commit. Encodes the definition of done: understand, implement with red-green-refactor for backend logic, run typecheck and tests until clean, then commit. Also the unit of work for autonomous loops.
Use when creating or revising any Claude Code skill, or when the user says "write me a skill for X". Encodes the house conventions: frontmatter style, body structure, when to bundle scripts, and project vs user scope.
| name | prd-to-backlog |
| description | Use when a PRD or spec exists and needs to become an executable backlog. Breaks the document into GitHub issues structured as vertical-slice tracer bullets, marks each as HITL or AFK, and files a final QA issue. Requires the gh CLI. |
Turn a PRD into GitHub issues that agents and humans can execute from the same board. Issues beat rigid phase documents: they reorder, they absorb QA findings as new items, and gh makes them agent-readable.
Read the PRD. Decompose into slices where each issue crosses every layer it touches (data, logic, route, UI) and is independently verifiable. Tracer bullets, per The Pragmatic Programmer (Thomas & Hunt).
Label every issue:
hitl: needs human taste. Design decisions, UX judgment, architecture choices, anything where the spec says "feels right".afk: spec is pinned down and feedback loops (tests, typecheck, screenshot) can verify the result without a human watching. These are eligible for autonomous loops.When unsure, mark hitl. A wrongly-AFK'd issue costs a bad merge; a wrongly-HITL'd one costs a few minutes of supervision.
gh label create hitl --color FFA500 --force
gh label create afk --color 0E8A16 --force
gh issue create --title "..." --label afk --body "..."
Each issue body contains: goal (one line), the PRD section it implements, acceptance criteria as a checklist, and verification command(s). Use blocking references ("Blocked by #1") to encode order.
Always file a last issue, labeled hitl: "QA pass: use the product like a user. File every flaw found as a new issue." This is where new backlog items come from after v1.
Report the issue list with numbers, labels, and the dependency chain. The backlog is now the queue for /ship sessions or a Ralph loop.