Documentation
Everything you need to read, without leaving the site.
These pages are rendered from the repository at build time, so they always match the release you would install. Every page links to its source file when you want the original.
Get started
- QuickstartThis guide starts at a fresh checkout and ends with your first successful rd-skills task. Run commands from the repository root.4 min read5 sections
- How it worksrd-skills chooses the professional knowledge and mechanisms needed for your4 min read6 sections
- Advanced Installation & Recovery ReferenceUse this reference for non-default paths, project or admin scope, previewing changes, conflicts, backups, upgrade, uninstall, OpenAI API packages, and recovery. For a normal first installation, use Qu6 min read8 sections
- UsageDescribe the engineering outcome you want. You do not need to choose a specialist,4 min read5 sections
Operating model
- Operating Modelsrc/ contains the control prompt, four Agent Profiles, three Skill layers,3 min read5 sections
- Subagent and Review ModelThe four Profile boundaries are defined in3 min read4 sections
- AI Control BoundariesThis document defines the authority, permission, evidence, and context boundaries2 min read4 sections
- Hookless ArchitectureStart with How it works for the user-facing mental model.1 min read
Authoring skills
- Skill Content Governancerd-skills Skills are written for AI execution. The always-loaded root must provide the smallest complete decision contract; detailed or low-frequency material belongs in targeted references.38 min read8 sections
- Quality Modelrd-skills quality is the combination of control correctness, professional judgment, bounded context, observable evidence, and maintainable packaging.2 min read6 sections
- Marketplace IndexThe marketplace index and catalog are local, source-derived discovery views.2 min read4 sections
Assurance
- ValidationEvery completion or release claim needs evidence from the final material edit.25 min read8 sections
- Benchmarksrd-skills uses bounded, reproducible repository evaluations. Benchmarking is6 min read6 sections
- ScorecardThis handwritten table defines release expectations. It is not generated2 min read
- ReleaseRelease only from authored source, current evidence, generated build output,7 min read8 sections
In the repository
Generated inventories and governance files are large or change with every release, so they stay where they are produced.