Audit and refine skill repositories using a 12-dimension scorecard.
npx skills add https://github.com/yknothing/skills-refiner --skill skills-refiner
Please help me install this Agent Skill. Skill: skills-refiner Source: https://github.com/yknothing/skills-refiner/tree/main/skills/skills-refiner Command: npx skills add https://github.com/yknothing/skills-refiner --skill skills-refiner
Audits and refines a skill repository, a single skill, a workflow framework, or an eval set. It covers design quality, context engineering, purpose fit, evidence discipline, and boundary clarity — the structural dimensions that assertion-based testing does not reach. When a target_repo is provided, it continues into compatibility review, extraction, and integration planning, complementing skill-creator by providing deep design-level judgment after functional tests pass.
Run skills-refiner on a finished skill to generate a Stage 1 audit and actionable refinement actions.
High-intent search queries and answers about installing and using this skill.
To audit a skill repository, use a structured 12-dimension scorecard that evaluates positioning, structure, governance, safety, and maintainability. This design audit reveals structural dimensions and boundary clarity that standard assertion-based testing cannot reach.
Integrating a skill into a workflow framework requires explicit integration planning after a compatibility review. You classify extraction findings into four adoption categories, then fit the refined skill into the destination repository or workflow.
You need a design audit when functional tests pass but you still need to evaluate context engineering, purpose fit, evidence discipline, and boundary clarity. It diagnoses structural quality issues that assertion-based testing does not cover.
Yes, you can extract and classify skills using a four-way extraction method. This classifies findings as directly adoptable, adoptable after redesign, valuable as a pattern, or should-not-adopt, guiding your compatibility planning and upgrade process.
Refine a single skill by running a Stage 1 audit to generate actionable refinement actions. It diagnoses the skill's structure, boundaries, and context engineering to guide targeted improvements across twelve governance and safety dimensions.
MadAppGangAnalyze codebases with AST analysis and PageRank to map architectural layers.
gwozaiGenerate visual art concepts guided by a design philosophy.
VanLengsGenerate initial skill implementations from design artifacts using skill-creator.
fdshg693Validate Copilot integration with a minimal YAML frontmatter Skill unit.
