Automate drafting, testing, and iterative refinement of AI agent skills.
npx skills add https://github.com/yuki-yano/dotfiles --skill skill-creator-yuki-yano
Please help me install this Agent Skill. Skill: skill-creator Source: https://github.com/yuki-yano/dotfiles/tree/main/.agents/skills/skill-creator Command: npx skills add https://github.com/yuki-yano/dotfiles --skill skill-creator-yuki-yano
This tool guides users through the full lifecycle of skill creation: from defining intent to drafting a first version, then evaluating it with deterministic prompts and agent backends, and finally improving the description and performance based on structured feedback.
Draft your skill concept, run an initial evaluation, and iterate using measured feedback until the skill reaches a robust, validated state.
High-intent search queries and answers about installing and using this skill.
Automating AI agent skill creation involves orchestrating drafting, testing, and iterative refinement with structured feedback. This process integrates with Claude or Codex backends to evaluate performance using deterministic prompts and qualitative metrics.
The best way to iteratively refine prompts is through a description optimization loop that applies structured evaluation metrics. This loop measures deterministic feedback from agent backends to improve triggering accuracy and overall skill reliability.
Yes, automated skill evaluation works directly with Claude and Codex backends. It uses these dependencies to run deterministic eval prompts, generate qualitative feedback, and validate performance throughout the skill lifecycle.
Improving skill triggering accuracy requires running a description optimization loop based on structured evaluation feedback. By measuring deterministic eval prompt results, you refine the skill description until it reaches a validated, robust state.
Automated skill evaluation relies on deterministic eval prompts and structured feedback from Claude or Codex backends. Limitations include the dependency on these specific environments and the need for measured iterative refinement rather than immediate validation.
Yes, you can draft and test AI skills end-to-end without manual evaluation by automating the full lifecycle. The process defines intent, drafts an initial version, and evaluates performance using agent backends with a viewer for qualitative feedback.
AiasCoordinate end-to-end Claude Code skill creation with evaluation loops and benchmark analysis.
accolverCreate and refine agent skills through scripted eval loops.
MisLinkDraft SKILL.md files, generate test prompts, and run with-skill versus baseline evaluations.
drgaciwEvaluate LLM applications using automated metrics, human feedback, and benchmarking.
