percept-summarize
由 jarvis563 提供
Transportation简介
Automatic conversation summaries with entity extraction and relationship mapping.
描述
该技能的功能
Automatically generates summaries after conversations end, triggered by 60 seconds of silence. Extracts entities like people, companies, and topics, identifies action items, and maps relationships between entities. Summaries are stored in SQLite and searchable via full-text search or a local dashboard.
Combines automatic transcription, entity resolution, and relationship graphing in one step rather than requiring separate tools for each.
何时使用
- Retrieving notes from a meeting that just ended
- Finding action items from a past conversation
- Searching who was mentioned in discussions last week
- Pulling context about a client from recent calls
- Asking an agent to recap what was discussed
示例工作流
User asks: 'What did we talk about in the last meeting?'
- 1
Detects 60 seconds of silence and triggers summarization
- 2
Builds a speaker-tagged transcript from the recorded conversation
- 3
Sends transcript to OpenClaw LLM for summarization and entity extraction
- 4
Stores summary, entities, and relationships in SQLite
- 5
Returns summary with action items and linked entities
A structured summary with extracted people, organizations, topics, action items, and relationship connections stored locally and ready to query
前置要求
- percept-listen skill installed and running
安装
clawhub install jarvis563/percept-summarize运行 CLI 命令,或把这段提示词粘贴到 Claude Code、Cursor、Codex 或你在用的任意编程智能体里。