source-to-skill
Installation
SKILL.md
Source to Skill
Use this skill when the user wants to turn source material into a reusable agent skill that helps future agents do better work, not merely remember a document.
The goal is to make generated skills maximally useful: narrow enough to trigger at the right time, dense enough to guide behavior, small enough to load cheaply, and structured enough for agents to retrieve details on demand.
Load agent-customization for skill frontmatter, discovery, routing, or customization-file mechanics; return to this workflow for source extraction, synthesis, provenance, and validation.
Read this first
Use this default safe path unless the user explicitly asks for analyze-only mode or the source fails a gate below:
- Confirm source access, destination, and intended use.
- Analyze the source into frameworks, decisions, workflows, examples, risks, and terms.
- Produce an analysis report first when source quality or scope is uncertain.
- Generate a compact
SKILL.mdplus optional reference files only when they remove real complexity. - Put the most important behavioral rules in
SKILL.md; put long details inreferences/orexamples/. - Add provenance for every major extracted concept.
- Do not leak local absolute paths, temporary extraction paths, workspace storage paths, or CLI argv into generated artifacts.
- Validate frontmatter, links, trigger specificity, and token budget before reporting success.