菜单
Use ONLY when the user explicitly asks for it — e.g. "create a product brief", "flesh this out", "turn this idea into a product brief", or similar direct requests. Do NOT trigger on general product discussions or idea sharing unless the user explicitly invokes this skill by name.
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 ONLY when the user explicitly asks for it — e.g. "explain this code," "walk me through this module/feature/API," "review this for me," "/explain-code," or points at a file/function/snippet and asks what it does. Produces a short, high-level report on existing (already-committed, no-diff) code: what it does, how it's structured, non-obvious behavior, edge cases, risk flags, open questions, and where to look next. Do NOT trigger automatically just because code is present in context or was recently discussed — wait for the user to ask.
Use ONLY when the user explicitly asks for it — e.g. "practice IELTS speaking", "IELTS speaking practice", "help me with IELTS", "IELTS part 1/2/3", or similar direct requests. Do NOT trigger on general mentions of IELTS or English speaking unless the user explicitly invokes this skill by name.
Use ONLY when the user explicitly asks for it — e.g. "mock interview me", "do a mock interview", "practice interview questions", "interview me", "drill technical Q&A", or similar direct requests. Do NOT trigger on general interview-related discussion unless the user explicitly invokes this skill by name.
Use ONLY when the user explicitly asks for it — e.g. "document this API", "hand off this endpoint", "prepare FE integration notes", or similar direct requests. Do NOT trigger automatically when documenting or summarizing code — wait for the user to ask.
Use ONLY when the user explicitly asks to enter brainstorming mode by name (e.g. "brainstorm first", "let's brainstorm", "start a brainstorming session"). Do NOT trigger on general discussion, questions, requests for suggestions, or solution exploration unless the user explicitly invokes this skill by name.
Use ONLY when the user explicitly asks for it — e.g. "give me the change report," "summarize the changes," "what did you change," "/change-report," or similar direct requests. Produces a short, high-level report (not a line-by-line walkthrough) covering what changed, behavior differences, assumptions made, known gaps, risk areas, and what to verify manually — so the user can review quickly without reading the whole diff. Do NOT trigger this automatically after writing or editing code — wait for the user to ask.
| name | bdv-product-brief |
| description | Use ONLY when the user explicitly asks for it — e.g. "create a product brief", "flesh this out", "turn this idea into a product brief", or similar direct requests. Do NOT trigger on general product discussions or idea sharing unless the user explicitly invokes this skill by name. |
Interview the user in 4 groups, one at a time. Don't force answers; unknowns go to Open Questions.
After the user's brainstorm, actively suggest additions they likely haven't considered: adjacent features, common patterns for this type of product, edge-case user needs, monetization angles, or risks. Mark these clearly as your suggestions, for example "Suggested by AI", so the user can accept or reject each one.
Then draft using assets/product-brief.md. Tag each feature, user's and
suggested, as Core if it directly solves the stated problem or Extension
if it is future or nice-to-have.
Push back on vague problem statements or feature lists that aren't grounded in a real problem.
Ask the user to review and iterate until they explicitly agree that the proposed Product Brief is ready.
After agreement, create docs/product/ if it does not exist, then write the
brief to docs/product/product-brief.md. Do this automatically without asking
for a destination first.
If docs/product/product-brief.md already exists, do not overwrite it. Report
the conflict to the user and ask for instructions, such as where to save the new
brief or what to do with the existing file.
After writing the file, report the saved path back to the user.
