The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-supplied data, such as application screenshots, screen captures, and code snippets, to perform design reviews and generate improvement briefs.
Ingestion points: User-provided artifacts (screens, code paths, and captures) processed in the review workflow defined in references/review-playbook.md and references/platform-foundations.md.
Boundary markers: The skill emphasizes a distinction between 'observation' (direct evidence from user data) and 'hypothesis' (inferred behavior), which serves as a logical boundary for the agent.
Capability inventory: The skill primarily generates structured documentation (briefs, findings) and delegates remediation tasks to other specialized skills (omh-frontend, omh-accessibility-audit). It uses custom commands like omh runtime record for state tracking.
Sanitization: While no technical sanitization is mentioned, the instructions explicitly prohibit treating prepared briefs as implementation or certification, limiting the impact of malicious content in ingested data.
[SAFE]: The skill references external documentation and libraries from established sources (e.g., Apple Developer portal, W3C, and GitHub repositories for GSAP and other UI tools). The instructions include defensive measures, such as explicit warnings against the automated installation of libraries or the use of runtime-fetched CDN scripts.

omh-apple-design