搜索

PythonReactSEOPDF
GitHub Skill

gen-plan

Generate iterative `plan-N.md` files in the repo root (max N=5). Use only when asked; clarify before `plan-1.md`; stop at `plan-5.md`.

tkersey☆ 37 仓库 Stars0 仓库 Forks

安装与使用

请根据来源仓库和你的 AI 工具说明安装此技能。先阅读技能文件和依赖要求,再添加到工作流。

查看来源 ↗使用指南

技能说明

Gen-Plan

Contract

  • Scope: operate in the repo root only; manage files named plan-N.md where N is an integer.
  • Output location is fixed: write plan-N.md to the current repo root directory only.
  • Never write plan.md, PLAN.md, or any non-plan-N.md filename.
  • Never write outside the repo root. Prohibited examples include ~/Downloads, $HOME, absolute paths, or sibling directories.
  • If any instruction or path suggests writing outside the repo root, stop and ask for clarification.
  • Define N as the maximum numeric suffix among files matching plan-(\d+).md (ignore non-matching filenames, including legacy plan-N-E.md / plan-N-R.md).
  • If any matching file has N > 5: do nothing; reply exactly: "Plan is ready."
  • If plan-5.md exists (case-insensitive): do nothing; reply exactly: "Plan is ready."
  • If no matching plan-(\d+).md exists: run the clarification flow, then create plan-1.md.
  • Otherwise: create plan-(N+1).md. Source plan markdown: plan-N.md.
  • When loading plan markdown from a source file, insert the full contents of plan-N.md at the <INCLUDE CONTENTS OF PLAN FILE> placeholder in the plan template.
  • Never overwrite an existing target file; stop and report the conflict.
  • Ask questions only when unresolved judgment calls block the next iteration.

Clarification flow (when needed)

  • Research first; ask only judgment-call questions.
  • Use the CLARIFICATION EXPERT: HUMAN INPUT REQUIRED block with numbered questions.
  • After questions are answered, determined if another round of questions is needed; if so continue asking questions until there are no more then write the next plan file.

Iterate on the plan

Purpose: Use the prompt below as an internal instruction to produce plan-(N+1).md from plan-N.md.

Output rules:

  • The plan file must contain the response to the prompt, not the prompt itself.
  • Do not include the prompt text, blockquote markers, or nested quotes in the plan file.
  • The output must be normal Markdown (no leading > on every line).
  • When inserting the source plan, include it verbatim with no extra quoting, indentation, or code fences.
Prompt template (verbatim, internal only — never write this into the plan file)

Carefully review this entire plan for me and come up with your best revisions in terms of better architecture, new features, changed features, etc. to make it better, more robust/reliable, more performant, more compelling/useful, etc.

For each proposed change, give me your detailed analysis and rationale/justification for why it would make the project better along with the git-diff style changes relative to the original markdown plan shown below:

使用示例

查看来源仓库中的使用示例与配置说明。

查看来源 ↗

来源历史

收录时间
2026-01-20
索引更新时间
2026-01-26
Branch
main

收录与同步日期属于本站记录。源码修改历史请查看来源仓库。

GitHub ↗

chat Comments (0)

chat_bubble_outline

No comments yet. Be the first to share your thoughts!