skillit

Generate structured AI agent skills (SKILL.md) and llms.txt from your TypeScript API documentation

When to Use

• You have a Commander program and want structured option/argument extraction with full fidelity
• Runtime introspection is unavailable (no access to the program object)
• The CLI uses a framework other than Commander (yargs, oclif, custom)
• You have both CLI surfaces (from introspection/help) and typed config interfaces (from TypeDoc)
• You want JSDoc @useWhen/@avoidWhen/@never tags to appear on CLI options in the generated skill
• You have a Commander program and want to generate a skill from its command structure
• You have raw --help output and no runtime access to the program object

Avoid when:

• Your CLI uses yargs, oclif, or another framework — use parseHelpOutput as a fallback instead
• Your CLI is built with a framework other than Commander — use parseHelpOutput directly instead
• API surface: 4 functions

NEVER

• NEVER pass both program and helpTexts — program takes precedence and helpTexts is silently ignored
• NEVER forget to pass configSurfaces when you have typed option interfaces — JSDoc metadata won't be correlated

Configuration

CliExtractionOptions

| Key | Type | Required | Default | Description | | ---------------- | ---------------------------------------------------------------------------------------------------- | -------- | ------- | -------------------------------------------------- | | program | any | no | — | Commander program object (preferred) | | helpTexts | Record<string, string> | no | — | Help text per command (fallback) | | metadata | { name?: string; description?: string; keywords?: string[]; repository?: string; author?: string } | no | — | Package metadata | | configSurfaces | ExtractedConfigSurface[] | no | — | Config surfaces from TypeDoc for JSDoc correlation |

Quick Reference

Commander: introspectCommander Fallback: parseHelpOutput Correlation: correlateFlags Extraction: extractCliSkill

Links

• Repository
• Author: Pradeep Mouli