菜单
Use FIRST when any new piece of work arrives — a request, feature, change, fix, question, or idea — before starting on it or choosing an approach. Not for continuing work already routed to a workflow skill.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
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。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/JetBrains/thinkrail --skill choosing-a-workflow命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
| name | choosing-a-workflow |
| description | Use FIRST when any new piece of work arrives — a request, feature, change, fix, question, or idea — before starting on it or choosing an approach. Not for continuing work already routed to a workflow skill. |
The root router of the workflow family in packages/pi-thinkrail-workflow. It classifies the incoming
work and names the workflow skill that governs it — nothing more. A routed skill's steps live in that
skill alone; read it, don't run it from memory.
Read the request and the workspace, then answer three questions — usually silently, from what is already in front of you:
If the route is genuinely ambiguous from the request alone, ask one short clarifying question
(ask_user_question, composed per the asking-user-questions concept skill) rather than guessing.
| Classification | Route |
|---|---|
| Project onboarding — no spec graph yet: an empty workspace with a raw idea, or an existing codebase to set up / spec | Read and follow setting-up-a-project — a dispatcher that routes on the workspace's state |
| The PR lifecycle — finished work to ship as a pull request, or an existing PR to tend: create, bring up to date, screenshots, checks, review comments | Read and follow shipping-a-pr |
| The work creates or changes anything in the project — a new feature, added functionality, a behavior change, a nontrivial design decision | Read and follow brainstorming — however small it looks; that skill owns the "too small" judgment |
| Anything else — answering questions, explaining code, running commands or checks, work that changes nothing | No matching workflow. Say so in one line (e.g. "No workflow skill covers this; proceeding directly.") and proceed with your own judgment. Never stretch a route to fit — a forced route is worse than none. |
One route per piece of work. If a request bundles work from different rows (e.g. "explain X, then change Y"), route the part that changes the project and handle the rest directly.
This skill ends by naming exactly one of: setting-up-a-project, shipping-a-pr, brainstorming, or no matching workflow (proceed with judgment). Adding a workflow to the family adds a row to the table above — see the writing-workflow-skills skill.
Use when a reflection package hands you another agent's review findings to verify (before they become a fix request): you are the REFLECTOR, an independent skeptic. Judge each finding against the real code and settle it with reflect_finding — kept or refuted.
Use when a review package asks you to review a plan step's change set (todo.startReview): you are the REVIEWER, not the author. How to judge an agent-written diff, file findings with add_review_comment, and settle with exactly one review_verdict.
This chat has a shared, live TODO plan — your tasks for the conversation, which the user also edits. Read this skill and reach for the todo_* tools whenever a request takes more than a couple of steps. It covers the plan model (group = task, items = its steps; loose items are the user's lane), how to work it: propose the plan FIRST (todo_write before you ask questions or start work), work tasks strictly in order with one step in_progress, keep statuses current, re-read the list (source of truth) to catch the user's edits, respect removals, and never delete done items.
