Evaluate and optimize agent skill files using an 8-dimension rubric.
npx skills add https://github.com/Herxinsasa/Skills-Collector --skill darwin-skill-herxinsasa
Please help me install this Agent Skill. Skill: darwin-skill Source: https://github.com/Herxinsasa/Skills-Collector/tree/main/darwin-skill Command: npx skills add https://github.com/Herxinsasa/Skills-Collector --skill darwin-skill-herxinsasa
Agent skill ecosystems expand rapidly, but manually maintaining dozens of skills is unsustainable. Traditional review only checks structural formatting, missing real-world effectiveness issues. This skill provides a systematic, automated way to evaluate and continuously improve agent skills using both static analysis and live testing.
Use the darwin-skill optimizer to evaluate and improve your agent skills by running the optimization loop on your skill collection.
High-intent search queries and answers about installing and using this skill.
Automate agent skill quality scoring by evaluating files against an 8-dimension rubric combining 60 points for static structure analysis and 40 points for live effectiveness testing.
Continuously improve prompt engineering outputs through hill-climbing optimization, which diagnoses the weakest scoring dimension and generates targeted improvements to systematically enhance skill files.
Hill climbing optimization for agent skills works by applying a git ratchet mechanism that permanently retains only score-increasing changes, pausing after each cycle for human confirmation before proceeding.
You can optimize large Claude Code skill collections by running a full optimization cycle that evaluates each skill against realistic test prompts using sub-agent execution for independent scoring.
You need git version control to run agent skill evaluation because the optimization process relies on a git ratchet mechanism to track ratchet-based improvements and preserve only beneficial changes.
Static structure analysis for agent skills is limited to checking frontmatter quality and workflow clarity, missing real-world effectiveness issues that require live testing with sub-agents to fully evaluate.
lxp119Optimizes Claude Code Skills using an 8-dimensional rubric and git version control with user confirmation.
buzzbysolcexRender BuzzBD-styled HTML templates into PNGs and upload them to Twitter.
dhamijaSimulate real users to test web and mobile UIs with LLM-driven protocols.
lioartoilOptimize headless PR review prompts via iterative hill-climbing loops on GitHub.
