transcript-analyzer
此技能使用Cerebras AI(llama-3.3-70b)分析会议记录,提取决策、行动项、观点、问题和术语。用于用户请求分析记录、从会议中提取行动项、在对话中找到决策、从讨论中构建术语表或总结记录会议的关键点。
📖 中文导读:此技能使用Cerebras AI(llama-3.3-70b)分析会议记录,提取决策、行动项、观点、问题和术语。用于用户请求分析记录、从会议中提取行动项、在对话中找到决策、从讨论中构建术语表或总结记录会议的关键点。 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。
name: transcript-analyzer description: This skill analyzes meeting transcripts to extract decisions, action items, opinions, questions, and terminology using Cerebras AI (llama-3.3-70b). Use this skill when the user asks to analyze a transcript, extract action items from meetings, find decisions in conversations, build glossaries from discussions, or summarize key points from recorded meetings.
Transcript Analyzer
Overview
Analyze meeting transcripts using AI to automatically extract and categorize:
- Decisions - Explicit agreements or choices made
- Action Items - Tasks assigned to people
- Opinions - Viewpoints expressed but not agreed upon
- Questions - Unresolved questions raised
- Terms - Domain-specific terminology for glossary
Prerequisites
Before first use, install dependencies:
cd ~/.claude/skills/transcript-analyzer/scripts && npm install
Usage
To analyze a transcript:
cd ~/.claude/skills/transcript-analyzer/scripts && npm run cli -- <transcript-file> -o <output.md> [options]
Options
| Option | Description |
|---|---|
<file> |
Transcript file to analyze (first positional arg) |
-o, --output <path> |
Write markdown to file instead of stdout |
--include-transcript |
Include full transcript in output [default: off] |
--no-extractions |
Exclude extractions section |
--no-glossary |
Exclude glossary section |
--glossary <path> |
Custom glossary JSON path |
--skip-glossary |
Don't preload glossary terms |
--max-terms <num> |
Limit glossary suggestions |
--chunk-size <num> |
Override chunk size (default: 3000) |
Examples
Basic Analysis
cd ~/.claude/skills/transcript-analyzer/scripts && npm run cli -- /path/to/meeting.md -o /path/to/analysis.md
Include Original Transcript
cd ~/.claude/skills/transcript-analyzer/scripts && npm run cli -- /path/to/meeting.md -o /path/to/analysis.md --include-transcript
Extractions Only (No Glossary)
cd ~/.claude/skills/transcript-analyzer/scripts && npm run cli -- /path/to/meeting.md -o /path/to/analysis.md --no-glossary
Analyze Specific Section
To analyze only part of a transcript, extract the section first:
sed -n '50,100p' /path/to/meeting.md > /tmp/section.md
cd ~/.claude/skills/transcript-analyzer/scripts && npm run cli -- /tmp/section.md -o /path/to/section-analysis.md
Output Format
The tool generates markdown with:
YAML Frontmatter - Processing metadata:
- chunks processed
- extractions count by type
- new terms discovered
- model used (llama-3.3-70b via Cerebras)
- token usage (input/output/total)
Extractions - Categorized findings with confidence scores:
- Each extraction includes speaker (if identified), source snippet, and related terms
Glossary - Approved terms from existing glossary + suggested new terms with definitions
Configuration
The skill uses Cerebras API with the key stored in scripts/.env:
CEREBRAS_API_KEY=<your-key>
Scripts
scripts/cli.ts- Main CLI entry pointscripts/src/lib/extract-service.ts- AI processing logic using Cerebrasscripts/src/lib/markdown.ts- Markdown output generationscripts/src/lib/term-utils.ts- Term deduplication utilitiesscripts/src/lib/mockExtractor.ts- Mock mode for testingscripts/src/types/index.ts- TypeScript type definitionsscripts/data/glossary.json- Default glossary storage