Automates eval-guided improvement of SKILL.md through audit and mutation.
npx skills add https://github.com/surahli123/autorefine-skill-improvement --skill autorefine
Please help me install this Agent Skill. Skill: autorefine Source: https://github.com/surahli123/autorefine-skill-improvement/tree/main/autorefine Command: npx skills add https://github.com/surahli123/autorefine-skill-improvement --skill autorefine
AutoRefine provides a guided, eval-grounded workflow to iteratively improve a SKILL.md by surfacing misalignments, validating improvements with judges and domain tests, and mutating toward verifiable gains.
Run AutoRefine on a target skill path to start the guided improvement workflow.
High-intent search queries and answers about installing and using this skill.
AutoRefine improves a SKILL.md by automating eval-grounded autoresearch, combining a design-audit with Hamel's Three Gulfs and a mutation loop to validate changes through judge validation and domain tests.
The mutation loop process for skill improvement applies phase-based mutations to a SKILL.md, routing downstream work through pattern-aware strategies and contract integrity checks to ensure verifiable, eval-grounded gains before session-close cleanup.
A design-audit using Hamel's Three Gulfs evaluates skill workflows by surfacing misalignments between intent and execution, providing structured preflight checks before entering the mutation loop to refine the SKILL.md.
Yes, you can apply automated evaluation and mutation to any skill defined by a SKILL.md, spanning automation to agent workflows, using adapter-domain evaluation and ambient learning to verify improvements across diverse domains.
The best way to validate SKILL.md improvements with domain tests is running a disciplined mutation loop that uses judge validation and domain-metric integration, ensuring changes are verifiable and backed by eval-grounded evidence.
Your skill workflow needs pattern-aware routing and contract integrity checks to maintain verifiable improvements during mutation, ensuring adapter-domain evaluation aligns with ambient learning and preventing structural breakdowns across the SKILL.md.
Binh230199Create and audit VS Code Agent Skill directories and SKILL.md files.
Danik911Extracts Langfuse Cloud API traces, observations, and metrics for debugging, telemetry analysis, and audit trails.
leary-pokenRun MLflow GenAI evaluate() to score agent outputs with configurable scorers.
vibeingingCreate and refine YiYi skills with automated evaluation loops.
