搜索

PythonReactSEOPDF
GitHub Skill

ppage

Captures session learnings, decisions, and context to a markdown file for future agent ramp-up. Use when user says "ppage", "page context", "save context", "capture learnings", or before ending a substantial work session.

kaushikgopal☆ 63 仓库 Stars6 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

Page Context

Synthesize the current session's important context, learnings, and state into a document that another LLM agent can use to immediately ramp up.

When to Use

  • User explicitly invokes with "ppage" or similar
  • Before ending a long or complex session
  • When switching context to different work
  • When user wants to preserve learnings

Process

  1. Reflect on the session: what was accomplished, what was learned, what decisions were made
  2. Identify key files that are central to the work
  3. Determine current state and logical next steps
  4. Note any open questions or blockers
  5. Capture constraints and preferences: Document any user preferences, project constraints, or environmental limitations discovered during the session
  6. Gather critical references: Note important external URLs, documentation, example data, or key values discovered
  7. Create Bootstrap Instructions: List the key files in priority order (most important first) with absolute paths so a future agent can read them to reach full context
  8. Generate a succinct but descriptive name for the session (2-4 words, kebab-case)
  9. Open TEMPLATE.md and follow it exactly
  10. Write output to .ai/tmp/{YYYY-MM-DD}-{succinct-name}.md

Output Format

Use TEMPLATE.md as the output structure.

Guidelines

  • Be concise: Target under 200 lines
  • Be actionable: File paths should be exact, next steps should be concrete
  • Be explicit: Don't assume the reading agent has any prior context
  • Capture decisions: Include rationale so decisions aren't re-litigated
  • Preserve constraints: User preferences and project constraints prevent re-litigation
  • Name meaningfully: The filename should indicate what the session was about
  • Bootstrap first: The ppage is a summary, not a full context dump. The Bootstrap Instructions tell the future agent which files to read to reach full context. Prioritize files (most important first), use absolute paths, and separate required vs optional files.
  • Omit empty sections: If a section has no content (e.g., no constraints discovered, no external references), omit it entirely rather than leaving placeholders

Create the .ai/tmp/ directory if it doesn't exist.

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
2026-02-04
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!