Create and refine Claude skills by structuring SKILL.md and running evaluations.
npx skills add https://github.com/junkisai/dotfiles --skill skill-creator-junkisai
Please help me install this Agent Skill. Skill: skill-creator Source: https://github.com/junkisai/dotfiles/tree/main/.claude/skills/skill-creator Command: npx skills add https://github.com/junkisai/dotfiles --skill skill-creator-junkisai
This Skill provides a structured, repeatable workflow for creating, evaluating, and iterating Claude skills, helping teams produce high-triggering, well-documented skill units with minimal ambiguity.
Follow the loop: draft a skill, run evals, observe results in the viewer, and iterate until the triggering performance stabilizes.
High-intent search queries and answers about installing and using this skill.
To create and iterate on Claude skills, you structure a SKILL.md file, run automated trigger evaluations, compare results against baselines, and refine descriptions until triggering performance stabilizes.
The best way to optimize skill triggering descriptions is to run post-hoc improvements on the SKILL.md frontmatter using bundled scripts, allowing you to enhance triggering accuracy and track benchmarking results.
Yes, reproducible skill development across teams is supported by coordinating deterministic experiment setup, executing trigger evaluations, and packaging outputs for consistent benchmarking comparisons.
Benchmarking works by executing automated trigger evaluations, comparing outputs with established baselines, and visualizing the results with an eval-viewer to guide further iteration on the skill.
Yes, you need Python dependencies including anthropic and pyyaml to execute the helper scripts that run evaluations, generate reports, and improve skill descriptions.
The workflow bundles reusable scripts, references, and assets to support deterministic tasks like generating reports, running evaluations, and improving trigger descriptions for knowledge reuse.
Norman-buryInstall Miniconda and configure a reproducible Python data-visualization environment.
timurgaleevMonitor post-deploy web apps for defects and performance regressions.
POWERFULMOVESValidate SKILL.md YAML frontmatter and Markdown structure for consistency.
clineCreate and iteratively improve AI agent skills through structured evaluation workflows.
