share-analytics

Share Model Matchmaker Analytics

OPTIONAL SKILL - This is for contributing data back to the community. You don't need this for personal optimization. Use the optimize-classifier skill instead for private, local tuning.

This skill helps you contribute sanitized analytics data back to the Model Matchmaker community to improve the classifier, while protecting your privacy. This is completely optional and not required for Model Matchmaker to work or improve based on your usage.

What This Skill Does

Note: You can optimize your classifier privately using the optimize-classifier skill. This "share-analytics" skill is only if you want to contribute aggregated data back to help improve Model Matchmaker for everyone.

1. Reads your local Model Matchmaker logs (~/.cursor/hooks/model-matchmaker.ndjson)
2. Sanitizes prompt snippets by removing project names, file paths, and personal details
3. Categorizes prompts using a standard 20-category taxonomy
4. Aggregates patterns to show where the classifier succeeded or failed
5. Shows you the sanitized report for review before you share it

Instructions for the AI

You are helping the user create a privacy-safe contribution report from their Model Matchmaker usage data. Follow these steps:

Step 1: Read the Log File

Read the NDJSON log file at ~/.cursor/hooks/model-matchmaker.ndjson. Each line is a JSON object with this structure:

{
  "event": "recommendation",
  "ts": "2026-03-06T15:44:39.040834",
  "conversation_id": "abc123",
  "generation_id": "gen456",
  "model": "claude-4-opus",
  "recommendation": "sonnet",
  "action": "ALLOW|BLOCK|OVERRIDE",
  "word_count": 25,
  "prompt_snippet": "First 40 chars of prompt"
}

or

{
  "event": "completion",
  "ts": "2026-03-06T15:45:12.123456",
  "conversation_id": "abc123",
  "generation_id": "gen456",
  "model": "claude-4-opus",
  "status": "completed|errored|aborted",
  "loop_count": 0
}

Step 2: Categorize Prompts