recall
Recall what was done on a specific project or topic from past AI coding sessions. Use when the user says "what did I do on [project]", "recall [topic]", "history of [project]", "when did I last work on [thing]" — but also when they reference prior work without asking for history explicitly: "didn't we already fix this", "what was that approach we tried", "why did we do it this way", "what happened to [file/feature]". Use PROACTIVELY before re-deriving a decision or re-solving a problem a past session likely settled — past conversations hold rationale and dead ends that never made it into git.
来源信息
- 仓库
- nicknisi/sessions
- 最近来源活动
- 2026年7月16日 17:30
- 检测到的 SKILL.md 语言
- 英语
- 星标
- 35
- 分支
- 6
安装方式
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
检查来源文件
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
正在显示 SKILL.md
SKILL.md
来源说明 · 只读预览- name
- recall
- description
- Recall what was done on a specific project or topic from past AI coding sessions. Use when the user says "what did I do on [project]", "recall [topic]", "history of [project]", "when did I last work on [thing]" — but also when they reference prior work without asking for history explicitly: "didn't we already fix this", "what was that approach we tried", "why did we do it this way", "what happened to [file/feature]". Use PROACTIVELY before re-deriving a decision or re-solving a problem a past session likely settled — past conversations hold rationale and dead ends that never made it into git.
- argument-hint
- project name or path
Recall past work on a specific project or topic.
## Steps
1. **Classify the question.** Each shape has a cheapest path:
- **Needle** — "what did we decide about X?" → search, then read just the matched exchange.
- **Arc** — "where did we leave off on X?" → digest the relevant session(s).
- **Artifact** — "what happened to this file?" → search with the `files` filter.
2. **Find candidate sessions.** Call `search_sessions`:
- Needle/arc: `query` = topic keywords; add `project` if a path was given; `limit`: 10.
- Artifact: `files`: [path or suffix], no `query` — results come back newest first.
Each result carries `messageHits` — the specific messages that matched (`index`, `role`, `snippet`). The snippets alone often answer a needle question.
3. **Digest the best 1-3 candidates.** Call `get_session_digest` with each candidate's `filePath`. One bounded call (~2k tokens) returns the session's arc: every genuine user turn paired with its exchange's final assistant reply, each with a message `index`. For "where did we leave off", the last exchange of the newest session's digest usually is the answer.
4. **Expand only what you need.** To read around a specific point, call `get_session_messages` with `offset` = a `messageHits[].index` (from search) or `exchanges[].index` (from digest) and a small `limit` (5-10). Never page a transcript from offset 0 when a hit or digest index is available.
5. **Summarize the history.** Write a chronological summary:
### Sessions on {project/topic}
For each relevant session:
- **{date}** ({tool}) — What was worked on, key decisions made, outcome
**Overall arc:** How the work evolved across sessions.
## Guidelines
- Budget discipline: digests are bounded, transcripts are not. Prefer one digest over five pages of messages; expand at most the exchanges you need.
- If a digest returns empty `exchanges` (no genuine human turns), fall back to `get_session_messages`.
- Order chronologically (oldest first) to show the arc of work.
- For vague date-range asks ("what was I doing last week?"), use `get_activity_digest` with a date range instead of search.
- Focus on decisions and outcomes, not implementation details.
- If there are many sessions, group by phase or milestone rather than listing each one.
在 GitHub 查看