Analyze SKILL.md files with an 8-dimension rubric and optimize via git-backed hill-climbing.
npx skills add https://github.com/pingqLIN/UniText --skill darwin-skill-pingqlin
Please help me install this Agent Skill. Skill: darwin-skill Source: https://github.com/pingqLIN/UniText/tree/main/runtime/skills/darwin-skill Command: npx skills add https://github.com/pingqLIN/UniText --skill darwin-skill-pingqlin
Darwin.skill tackles the challenge of optimizing and vetting AI agent Skills by combining structural evaluation with live, test-driven feedback, ensuring improvements are measurable and auditable.
Install this skill and run the optimizer on your target skills to begin the evaluation loop.
High-intent search queries and answers about installing and using this skill.
To optimize AI agent skills, you run test prompts comparing Skill-enabled versus baseline performance. This process uses an 8-dimension rubric and a git-backed ratchet mechanism to keep improvements and revert regressions.
Git-backed hill-climbing optimization for skill evaluation is a ratchet mechanism that commits only proven improvements to your skills. It automatically reverts any regressions identified during live testing with test prompts.
An independent sub-agent prevents self-reporting bias in skill testing by separately evaluating the outputs generated by the agent. This ensures structural scoring and real-world effectiveness assessments remain objective and measurable.
Yes, you can use human-in-the-loop checkpoints to vet AI skill improvements. The optimizer requires human confirmation at key milestones before applying any changes, ensuring all modifications are auditable and manually approved.
Yes, you need git to run structural scoring and live testing on agent skills. The optimizer relies on a git-backed ratchet mechanism to commit approved improvements, revert regressions, and generate visual result cards for each change.
The 8-dimension rubric used for evaluating SKILL.md files is a static structural scoring framework combined with phase-appropriate test prompts. It measures real-world effectiveness to ensure skill improvements are measurable and auditable.
tangzheng202202Integrate Darwin.skill evaluation with skill-manager to scan, evaluate, evolve, and sync Hermes Skills.
LIMITLESS-LONGEVITYDetect and merge upstream updates for installed skill branches.
HerxinsasaEvaluate and optimize agent skill files using an 8-dimension rubric.
zhizhunbaoValidate and correct Markdown formatting, syntax, and structure with Python scripts and linters.
