‹ 首页

pin-wrap-up

@toongri · 收录于 昨天 · 上游提交 昨天

Use at the end of a work session to review the WHOLE session and record entities worth pinning. This is the manual, deliberate complete-sweep review — NOT an automated nudge. Triggers on "wrap up", "wrap-up", "session wrap", "end of session", "what should I pin".

适合你,如果每次工作结束都需要理清重点

/ 通过 npx 安装 校验哈希
npx oh-my-skill add toongri/oh-my-toong-playground/pin-wrap-up
/ 通过 bash 安装
curl -fsSL https://oh-my-skill.com/install.sh | bash -s -- toongri/oh-my-toong-playground/pin-wrap-up
/ 已经装过?验证本机副本,不用重装
npx oh-my-skill verify toongri/oh-my-toong-playground/pin-wrap-up
安装目标可用 --agent / --scope 或 --to 明确指定;省略时只会在唯一已存在的 agent 目录上自动选择,零命中或多命中会停止并提示。content_hash 缺失或不一致均拒装。
25GitHub stars
~687最小装载
~687含声明引用
~1.1K文本包总量
索引托管

怎么用

商店整理自技能原文 · 版本 731a80a · 表述以原文为准
它做什么

在工作会话结束时,Claude会手动回顾整个会话,识别并列出值得记录的实体(如外部资源、决定、架构事实等),让用户确认后通过脚本记录固定。

什么时候触发

当用户在工作会话结束时说出“wrap up”、“wrap-up”、“session wrap”、“end of session”或“what should I pin”时触发。

装好后可以这样说
Claude会扫描整个会话并列出候选实体。
Claude会应用固定标准筛选候选实体。
Claude会列出候选实体并等待确认。
技能原文 SKILL.md作者撰写 · MIT · 731a80a

pin-wrap-up

Conduct a whole-session review to identify entities worth recording in the pins knowledge graph.

This is a deliberate, manual sweep. It is NOT automatic. You invoke it when a session is winding down and you want to surface knowledge before context is lost.

Process
1. Sweep the session

Scan back over the full session — every discovery, decision, external reference, code location, person named, and architectural fact that surfaced. Ask: "Would a future me (or a colleague) benefit from being able to find this in 10 seconds?"

Apply the recording rubric from the pin-record skill, using the worth/not-worth table below to decide what makes the cut.

Discard noise: transient commands, intermediate debug outputs, things that are already obvious from the codebase.

2. Collect candidates

List entities you propose to record. For each, briefly state:

  • What it is (one sentence)
  • Why it merits a pin (which axiom it satisfies)

Show this list to the user before recording. Confirm or prune before writing anything.

3. Validate, then record confirmed entities

For each confirmed candidate, pipe the JSON-serialized entity to the deployed script:

printf '%s' "$ENTITY_JSON" | bun "${CLAUDE_SKILL_DIR}/scripts/wrap-up.ts"

The script validates first. On FAIL it prints {"valid":false,"reason":"...","message":"..."} and stops — record is never called. On PASS it records and prints {"id":"...","status":"recorded"}.

Build the entity with the required body sections and complete frontmatter — see the pin-record skill for the body structure and full field reference.

If the script exits 0 and prints "status":"recorded", the pin was written.

4. Report

After recording, state:

  • How many entities were recorded
  • Their IDs and one-line summaries
  • Any that failed pre-validation (validate() returned valid: false) with the reason, so the user can decide whether to fix and retry

If the manifest is git-managed (git: true), commit the recorded pin file(s).

What belongs in a session wrap-up

| Worth pinning | Not worth pinning | |---------------|-------------------| | External SSOT located (URL, file, person) | Transient debug output | | Ground truth pinned in code (file:line) | Things obvious from reading the code | | Decision made that will constrain future work | Intermediate hypothesis that was disproved | | Person named as authority on a topic | Generic "check the docs" type notes | | Architecture fact that took effort to find | Information that will be stale within hours |

Constraint

Do NOT record everything. A knowledge graph full of low-value entries degrades retrieval. Quality over quantity — 2 high-quality pins beat 10 noisy ones.

按 MIT 许可原样转载,未经改动 · 在 GitHub 查看 →

评论

登录即可评论;带「已验证安装」的,是发布者名下有本店的安装或持有记录。