[FYI] Automated, natural language Skillopt skill

I built a skill (/skill-opt) to automate the entire evaluation, reflection, and patch deployment cycle through an interactive, natural language conversation with your agent.

Features

• Generates a self-contained Python optimizer on the fly using standard libraries (no pip install or external repo needed).
• Probes recent session transcripts to turn real developer corrections into regression benchmarks.
• Optimize a single file or point it at an entire folder of skills to tune them all in sequence.
• Can run on a recurring schedule to continuously optimize skills against recent real-life agent runs.
• Only commits patches with measured score gains on held-out tasks.
• Runs across Gemini, Claude, OpenAI, or OpenRouter with live progress updates in chat.
• Saves timestamped backups before modifying source files.

LMKWYT!

Start with the linked /skill-opt skill and read its description alongside the repository overview. The issue presents an external optimizer and lists its capabilities, but does not identify a requested change, target file, test, or acceptance condition, so a contributor should clarify the intended outcome before beginning work.