Reference
Skills
Host skill surfaces
Section titled “Host skill surfaces”Every host ships the same eleven domain knowledge skills. Claude and Cursor additionally own 17 slash commands. Codex exposes native explicit parity for those workflows as 17 command-workflow skills, for 28 Codex skills total.
| Domain skill | Content |
|---|---|
using-rn-dev-agent | Intent routing and artifact-first operating protocol |
rn-setup | Passive recovery diagnosis and consent-based setup routing |
rn-feature-development | Eight-phase feature implementation pipeline |
rn-testing | Maestro/action patterns, assertions, and persistent tests |
creating-actions | Reusable-action authoring, metadata, replay, and repair |
rn-debugging | Evidence-first screen/crash diagnosis |
| Device Control | Device lifecycle, screenshots, Expo/EAS, interaction tiers |
capturing-proof | Rehearsed PR evidence and strict receipt protocol |
sending-feedback | Reviewed and sanitized issue submission |
| Best Practices | 118 indexed RN/React rules and packaged checks |
rn-workflow | Proven operating sequence before a real device journey |
Domain skills may load from natural-language intent. Codex workflow skills are
explicit-only and use $rn-dev-agent:<workflow> [request text]; their
allow_implicit_invocation policy is false to avoid overlap with domain skills.
See the workflow index.