Automate end-to-end creation, evaluation, and packaging of Claude Skills.
npx skills add https://github.com/dEitY719/dotfiles --skill skill-creator-deity719
Please help me install this Agent Skill. Skill: skill:creator Source: https://github.com/dEitY719/dotfiles/tree/main/claude/skills/skill-create Command: npx skills add https://github.com/dEitY719/dotfiles --skill skill-creator-deity719
Skill creation and optimization is a multi-step, iterative process that benefits from a repeatable workflow. This Skill provides a structured path to capture intent, draft SKILL.md, define and run evals, benchmark results, and optimize triggering descriptions, enabling consistent, repeatable skill development.
Create a new skill by filling out SKILL.md and running the builder to test and package it.
High-intent search queries and answers about installing and using this skill.
You can automate Claude Skill creation by defining your intent in SKILL.md and running an automated eval-improve loop that drafts the skill, runs tests with train/test splits, and packages the result into a distributable .skill file.
Progressive disclosure is a technique used to structure skill descriptions so relevant information is revealed contextually, optimizing triggering accuracy and ensuring the skill activates only when appropriate for the user's request.
You benchmark Claude Skills by running the evaluation loop to generate benchmark.json and benchmark.md files, which integrate with the eval viewer to provide quantitative comparisons of performance across iterations.
YAML is required as a dependency to configure and structure the skill packaging workflow, allowing you to define references, scripts, and assets needed for the automated evaluation and improvement loops.
To improve an existing skill's triggering accuracy, use the iterative improvement workflow to analyze eval outcomes, refine prompts and tests, and re-run loops until the benchmark criteria are met.
Avoid automated skill creation workflows if your task requires a simple, single-step prompt without iterative testing, as the overhead of setting up train/test splits and benchmarking provides little value for non-complex skill definitions.
Cissou34730Generate SKILL.md skeletons and starter resources for Copilot skills.
hackfest-devCreate, test, and refine Claude skills with evaluation loops and frontmatter validation.
qiushibangCreate, modify, and improve AI agent skills through eval-driven workflows.
loongclaw-aiAutomate Claude skill creation with intent interviews, draft writing, and evaluation loops.
