GitHub - ttunguz/meeting-review-agent: AI-powered meeting performance review agent with weighted rubrics & improvement recommendations · GitHub

Source: original

Meeting Review Agent

An AI-powered daily meeting analysis tool that fetches meetings from Fireflies, clusters them by type, grades them using weighted rubrics, & generates actionable improvement recommendations.

Features

• Automatic Meeting Clustering : Categorizes meetings into Internal, 1:1, Pitch, Interview, or Other
• Weighted Grading Rubrics : Each meeting type has specific criteria with importance weights
• AI-Powered Analysis : Uses an LLM to evaluate transcripts against rubrics
• Daily Reports : Generates markdown reports with scores & improvement recommendations
• Scheduled Execution : Runs daily via launchd (macOS) or cron

Meeting Types & Rubrics

Internal Meetings

• Agenda clarity (20%)
• Decision outcomes (25%)
• Action items (25%)
• Time efficiency (15%)
• Participation balance (15%)

1:1 Meetings

• Career development (25%)
• Feedback exchange (25%)
• Blockers addressed (20%)
• Relationship building (15%)
• Action items (15%)

Pitch Meetings

• Problem understanding (15%)
• Solution clarity (15%)
• Market opportunity (10%)
• Team assessment (15%)
• Traction evidence (10%)
• Listening quality (10%)
• Question depth (10%)
• Reflection quality (5%)
• Immediate value delivery (5%)
• Relevance mapping (5%)
• Insight capture (5%)
• Red flag detection (5%)
• Next steps clarity (5%)

Interview Meetings

• Role clarity (15%)
• Skill assessment (25%)
• Culture fit (20%)
• Candidate engagement (15%)
• Structured evaluation (15%)
• Next steps clarity (10%)

Usage

Run for today

ruby meeting_review_daily.rb

Run for specific date

ruby meeting_review_daily.rb 2026-01-27

Configuration

Before running, update the following in meeting_review_daily.rb :

1. YOUR_NAMES : Set to your name variations for meeting filtering
2. INTERNAL_PARTICIPANTS : Set to your internal team members
3. API paths : Update the EXECUTE_PATH to point to your Code Mode executor (or replace with direct API calls)

Output

Reports are saved to reports/YYYY-MM-DD-meeting-review.md with :

• Executive summary (meeting counts by type)
• Daily weighted average score
• Per-meeting scores by criterion
• Prioritized improvement recommendations

Dependencies

• Ruby
• Fireflies API access (for meeting transcripts)
• An LLM API (Gemini, Claude, or similar) for transcript analysis

Scheduled Execution

Example launchd plist (macOS) :

Label com.example.meeting-review ProgramArguments /usr/bin/ruby /path/to/meeting_review_daily.rb StartCalendarInterval Hour 20 Minute 0

License

MIT