apple-design

Pass

Audited by Gen Agent Trust Hub on Sep 21, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill includes a feature to fetch live documentation from official Apple developer pages, which introduces a standard surface for indirect prompt injection via external data ingestion.
  • Ingestion points: The SKILL.md (Branch C) and references/hig-sitemap.md files provide a directory of official developer.apple.com URLs for the agent to retrieve live updates.
  • Boundary markers: The skill provides instructional guidance in SKILL.md to paraphrase content rather than reproduce it verbatim, which acts as a qualitative boundary, though no technical delimiters are enforced on the fetched data.
  • Capability inventory: The skill leverages common agent tools such as WebFetch or browser-based navigation to access external documentation.
  • Sanitization: The skill relies on the trusted nature of the official Apple documentation source and the agent's internal safety filters for sanitization.
  • [EXTERNAL_DOWNLOADS]: The README.md provides installation instructions using the npx skills utility and the claude plugin system. These are standard platform-supported methods for skill and plugin management. The external resources referenced are official repositories and tools associated with the skill's author and the supported platforms.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 21, 2026, 07:18 AM
Security Audit — agent-trust-hub — apple-design