agentleFS
coyvalyss1coyvalyss1/model-matchmaker/skills/share-analytics/SKILL.md

share-analytics

Skill168 starsChanged 8 months agoMITView on GitHub

Build a sanitized analytics report from your Model Matchmaker usage for community contribution. Optional; use optimize-classifier for private tuning instead.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 4 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read the NDJSON log file at ~/.cursor/hooks/model-matchmaker.ndjson.

    line 28
  2. Very specific:

    If a prompt doesn't fit any category, use other.

    line 88
  3. Distinctive:

    Remove all project/product names (PaymentService, Ross, DoMoreWorld, etc.)

    line 102
  4. Distinctive:

    Remove file paths and code references

    line 103
  5. Distinctive:

    Remove user names and email addresses

    line 104
  6. Distinctive:

    Keep the task category and generic action (e.g., "layout fix", "API integration", "refactor")

    line 105

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

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

Classify each recommendation event's prompt_snippet into one of these 20 categories. Never output the literal prompt_snippet — only the category and a generic pattern description.

High Frequency Categories (prioritize for training):

  1. ui_ux_bug_fixes - Layout, responsive design, overlap, positioning, visual glitches
  2. feature_implementation - New UI elements, components, flows, capabilities
  3. ai_chat_debugging - AI chat failures, model config, persona behavior issues
  4. data_persistence - Likes, votes, profile data, Firestore behavior
  5. testing_qa - Manual testing, browser automation, test flows

Medium Frequency Categories: 6. cross_platform_alignment - Bringing web and iOS features/design in sync 7. build_deployment - Xcode, build errors, deployment, environment setup 8. product_strategy - Strategic questions about product, market, differentiation 9. feature_design_ux - Designing flows, UX, feature behavior before implementation 10. marketing_content - Social posts, launch content, marketing copy 11. business_planning - Partnerships, events, proposals, business documents 12. documentation_process - Session logs, TODOs, internal docs 13. api_service_integration - Third-party APIs, image generation, external services 14. pricing_monetization - Plans, paywalls, pricing logic 15. plan_execution - Implementing predefined plans with todos

Low Frequency Categories: 16. platform_architecture - Tech stack, architecture, platform choices 17. brand_legal - DBA, trademarks, brand structure 18. configuration_oauth - Firebase, OAuth, auth setup 19. tool_troubleshooting - Non-code tools and environment issues 20. rd_competitive_analysis - Researching competitors, tech options, cost comparisons

If a prompt doesn't fit any category, use other.

Step 3: Sanitize Prompt Snippets

For each prompt snippet, extract the task type without revealing specifics:

Examples of sanitization:

  • "build the PaymentService checkout flow" → Category: feature_implementation, Pattern: "multi-component feature build"
  • "Fix the profile page overlap issue on mobile" → Category: ui_ux_bug_fixes, Pattern: "mobile responsive layout fix"
  • "Why is Ross not generating images like before?" → Category: ai_chat_debugging, Pattern: "AI persona behavior regression"
  • "git commit all changes" → Category: git_operation (if you add it, or use other), Pattern: "git commit"

Rules for sanitization:

  • Remove all project/product names (PaymentService, Ross, DoMoreWorld, etc.)
  • Remove file paths and code references
  • Remove user names and email addresses
  • Keep the task category and generic action (e.g., "layout fix", "API integration", "refactor")

Step 4: Correlate Recommendations with Completions

Match recommendation events with completion events using conversation_id. Calculate:

  • How many recommendations led to completed vs errored outcomes
  • Override rate per category
  • Error rate when user overrode vs. followed the recommendation

Step 5: Generate Contribution Report

Output a structured report in this format:

# Model Matchmaker Analytics Contribution

**Version:** [Model Matchmaker version from repo]
**Period:** [Date range from logs]
**Total Events:** [N recommendations, M completions]

## Summary Statistics

- Total Recommendations: N
- Allow: X (Y%)
- Block: X (Y%)
- Override: X (Y%)

## Override Analysis by Category

### High-Frequency Categories

**Category: ui_ux_bug_fixes**
- Override count: 12
- Model used → Recommended:
  - opus → sonnet: 8 times
  - sonnet → opus: 4 times
- Generic patterns:
  - "Responsive layout fixes classified as sonnet-level"
  - "Visual positioning bugs"
- Completion outcomes:
  - Completed after override: 10
  - Errored after override: 2

[Repeat for each high-frequency category with overrides]

### Medium-Frequency Categories

[Same structure]

### Low-Frequency Categories

[Same structure]

## Suggested Improvements

Based on the override patterns, suggest:
1. Keywords to add to classifier patterns (generic terms only)
2. Categories that need better model routing
3. Prompt characteristics the classifier missed

## What's NOT in this report

- No literal prompt text
- No project/product names
- No conversation IDs or timestamps
- No user identity
- All data has been generalized to task patterns

Step 6: Present for User Review

Show the user the full sanitized report and ask:

"This is your privacy-safe contribution report. Review it to make sure nothing sensitive is revealed. If it looks good, you can copy and paste this into the Model Matchmaker GitHub Discussions (or share however you like). Would you like me to make any changes?"


How to Use This Skill

As a user, invoke this skill by saying something like:

  • "Generate my Model Matchmaker analytics contribution"
  • "Create a sanitized report from my Model Matchmaker logs"
  • "Prepare my Model Matchmaker data for sharing"

The AI will read your logs, sanitize them, and show you the report for approval before you share it anywhere.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.