skill-factory

Skill-factory analyzes transcripts and workflow patterns to propose 3-5 candidate skills. Run `voyager factory propose` to mine observed tasks and generate skill proposals tailored to your automation needs, reducing manual scaffolding work.

Yes. After proposing candidates, use `voyager factory scaffold --name <skill>` to generate a complete SKILL.md with frontmatter validation, references, and prompts. The tool prevents overwrites without confirmation and keeps documentation under 100 lines.

Skill-factory maintains two indices: `generated_skills_index.json` tracks all scaffolded skills and `skill_proposals.json` stores proposal history. This ensures duplicate detection, audit trails, and organized skill inventory across coding workflows and PR review processes.

The tool detects duplicate proposals by comparing against existing indices before scaffolding. It requires explicit confirmation to overwrite existing skills and validates YAML frontmatter to ensure consistency across your skill collection.

Skill-factory mines coding workflows, PR review processes, and recurring automation tasks from transcripts. It identifies repeatable patterns in Claude Code interactions and uses LLM analysis to extract generalizable logic into standalone, reusable skills.