skill miner

The problem

Your agent history is a goldmine of patterns you do not notice. You repeat the same instructions, corrections, and context every day, wasting tokens and time re-explaining your preferences.

What it does

• Scans local history from over 20 different AI agents to find recurring requests.
• Redacts sensitive data including credentials and emails before analysis.
• Clusters thousands of requests into distinct tasks using lexical analysis and human-in-the-loop scoring.
• Generates three ready-to-install .skill files based on your specific historical preferences.
• Produces a detailed report on three additional automation candidates for future development.

Frameworks & tools

Python 3, SQLite, Claude Code, Cursor, and various local agent storage formats.

Why this beats prompting it yourself

Manual prompts are based on what you remember, which is often biased toward recent tasks. This tool uses cold data to identify high-friction repetitions you have become blind to, ensuring your skills handle the specific edge cases you have historically encountered.

Use cases

• Bootstrap a personalized skill library based on your actual work habits.
• Audit a team's agent usage to identify common bottlenecks.
• Turn a complex, multi-step workflow into a single, standardized agent command.
• Migrate your working preferences between different AI tools and agents.

Known limitations

Analysis requires at least 40 requests to be meaningful. It cannot access cloud-only history that is not mirrored to local disk.