ppage

将会话学习、决策和上下文保存到Markdown文件中,以供未来代理快速熟悉。当用户说“ppage”、“page context”、“save context”、“capture learnings”或结束一个实质性工作会话之前使用。

📖 中文导读:将会话学习、决策和上下文保存到Markdown文件中,以供未来代理快速熟悉。当用户说“ppage”、“page context”、“save context”、“capture learnings”或结束一个实质性工作会话之前使用。 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。

name: ppage description: 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. compatibility: Requires filesystem write access to create .ai/tmp/*.md in the current repository.

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.