AgentSkillsCN

index

扫描 .stoobz 目录中的会话存档,构建一份可搜索的过往工作索引。当用户输入“/index”、“查找上次的会话”、“列出我的会话”、“我之前都做了些什么”、“那次调查是在哪里开展的”,或需要按主题或工单定位过往会话时,此功能便可轻松帮您完成任务。它会遍历 .stoobz 目录,搜寻 TLDR.md、CONTEXT_FOR_NEXT_SESSION.md 等各类存档文件。

SKILL.md
--- frontmatter
name: index
description: Scan .stoobz directories for session artifacts and build a searchable index of past work. Use when the user says "/index", "find that session", "list my sessions", "what did I work on", "where was that investigation", or needs to locate a past session by topic or ticket. Scans for TLDR.md, CONTEXT_FOR_NEXT_SESSION.md, and other artifacts across .stoobz directories.

Index

Find and catalog past sessions across .stoobz/ directories.

Process

  1. Determine search scope:

    • Default: scan the nearest .stoobz/ directory (project-level first, then ~/.stoobz/)
    • If user specifies "all" or "everything": scan both project and global .stoobz/
    • If user provides a search term: filter results by topic/ticket
  2. Scan for session directories — Walk .stoobz/ looking for directories containing session artifacts (TLDR.md, CONTEXT_FOR_NEXT_SESSION.md, RETRO.md, PROMPT_LAB.md, HANDOFF.md).

  3. For each directory found:

    • Read the first 5 lines of TLDR.md (if present) for the title and date
    • Note which artifacts exist
    • Note the most recent modification date
  4. Present the index:

markdown
## Session Index — {scope}

| Directory                     | Date       | Summary                         | Artifacts |
| ----------------------------- | ---------- | ------------------------------- | --------- |
| `ENG-22456/uat-investigation` | 2026-02-06 | Statusline + session kit skills | T R P     |
| `ENG-22456/qa-phase`          | 2026-01-28 | QA environment profiling        | T C       |
| `memory-leaks`                | 2026-01-26 | BEAM memory investigation       | T C R     |

**Legend:** T=TLDR C=Context R=Retro P=Prompt-Lab H=Handoff
  1. If user is searching, highlight matching results and show context from the TLDR.

Rules

  • Read only headers — Don't load full file contents. First 5 lines of TLDR.md is enough for the index.
  • Sort by date — Most recent first.
  • Fast — This is a lookup tool. Don't analyze, just catalog.
  • Suggest pickup — If a result looks like it has unfinished work (has CONTEXT_FOR_NEXT_SESSION.md), note: "Has resume context — run /pickup from that directory."
  • Present to the user directly, don't write a file (this is a query, not an artifact).