Audit a skill's SKILL.md against canonical best-practices criteria.
npx skills add https://github.com/d3ming/dot-agents --skill improve-skill-d3ming
Please help me install this Agent Skill. Skill: improve-skill Source: https://github.com/d3ming/dot-agents/tree/main/master/skills/improve-skill Command: npx skills add https://github.com/d3ming/dot-agents --skill improve-skill-d3ming
Audits and improves an existing skill against canonical best-practices criteria.
Provide the target skill path or name as an argument to initiate an audit and improvement cycle.
High-intent search queries and answers about installing and using this skill.
You can audit and improve a Claude skill by evaluating its SKILL.md and adjacent references against a best-practices rubric, processing the diff to generate actionable recommendations and a backfill plan.
A diff-based skill evaluation isolates changes made during an improvement cycle, comparing the original skill files against modified versions to produce a targeted verdict and validate updates via a self-eval script.
Run the evaluation script against the target skill's repository files to trigger a self-eval, which validates the applied minimal changes and confirms whether the skill meets the best-practices criteria.
Yes, guided backfill applies minimal changes to the skill's frontmatter and body while preserving the original workflows, using a one-shot diff-based assessment to isolate and review modifications.
You need access to the target skill's repository files, including the SKILL.md, adjacent references, the run-eval script, and evaluation prompts to produce a diff-based verdict and backfill plan.
A failed verdict indicates the skill's frontmatter or body does not meet the canonical best-practices criteria, requiring you to apply the generated backfill plan and re-run the self-eval to validate improvements.
Animism001Automates creation, evaluation, refinement, and packaging of AI skills via SKILL.md and .skill files.
COCOSiL-incCreate and optimize Claude Skills with trigger evaluations and SKILL.md refinement.
simondhallidayEvaluate MCQs against a standardized rubric and identify failure modes.
konflux-ciEvaluate code changes for correctness, security, and style adherence.
