Guide interviews to draft SKILL.md and run iterative evaluation loops.
npx skills add https://github.com/mattstruble/skills --skill skill-creator-mattstruble
Please help me install this Agent Skill. Skill: skill-creator Source: https://github.com/mattstruble/skills/tree/main/skill-creator Command: npx skills add https://github.com/mattstruble/skills --skill skill-creator-mattstruble
The Skill Creator guides users in turning a vague idea for a new AI skill into a concrete, testable entry by capturing intent, drafting a structured SKILL.md, and outlining an iterative evaluation loop to drive improvement.
Draft a new skill from a user goal, run an initial eval loop, and iteratively refine the SKILL.md until the benchmarks improve.
High-intent search queries and answers about installing and using this skill.
To create an AI skill, you capture the initial intent through guided interviews to draft a structured SKILL.md file, then iteratively refine it through an evaluation loop until it meets benchmark standards. This structured approach turns abstract concepts into concrete, testable skills.
Evaluating prompt-engineering skills involves designing specific test prompts, running the agent with access to the skill, and collecting quantitative metrics. This guided evaluation loop translates benchmarking results into actionable improvements for iterative refinement.
Improving trigger accuracy requires description optimization, where you refine the triggering description in the SKILL.md file. By running iterative evaluations and analyzing benchmarking results, you can harden the skill to ensure concise, trigger-accurate entries.
Yes, guided interviews are used to capture user intent and draft the initial SKILL.md documentation. This process clearly defines trigger conditions and the skill's purpose before moving into the testing and iterative evaluation phases.
No, iterative refinement does not require separate tools because the creation and evaluation processes are integrated. You draft, run evals, and rewrite the skill within a continuous loop, expanding references or tests as needed to optimize performance.
Crispy13Draft SKILL.md files with frontmatter, instructions, and test workflows.
FxHollowMonitor deployed URLs for console errors, performance regressions, and page failures.
slowman2084Analyze prompt histories to detect anti-patterns and generate retrospective reports.
Shockwave2kDraft SKILL.md files and run evaluation loops to measure trigger accuracy.
