About this skill

Problem

Raw meeting notes often arrive as spoken transcripts: speaker switches, repetition, digressions, and implied context. When shared with a team, decisions, owners, and deadlines can get buried; manual cleanup requires repeatedly checking the source material. This skill targets the step from raw meeting text to a deliverable minutes document, not transcription or meeting management.

How It Works

Input can be meeting notes, transcript text, or note files provided by the user. The skill runs scripts/generate_minutes.py to parse the text and produce structured Markdown minutes. Key rules include:

Infer missing information: when the date, attendees, or topic are not explicit, fill them from context instead of leaving large gaps.
Distill rather than repeat: compress long exchanges while preserving decision logic, disagreements, and necessary rationale.
Make action items specific: attach an owner and due date where possible; otherwise mark them as “to be confirmed”.
Separate facts from opinions: present decisions and action items as facts, and label dissenting views as “some attendees believed” to avoid turning subjective comments into conclusions.

Boundaries

Default output is Markdown, useful for pasting into docs, wikis, or repositories. On request, it can create Tencent Docs or Word output, but that depends on external tooling. It does not replace recording, permissions, task systems, or long-term knowledge curation; if the input lacks key context, the result still needs human review.