apple-appicon
Pass
Audited by Gen Agent Trust Hub on Sep 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to visually inspect user-provided images for design compliance. While this creates a surface for potential injection via image content (OCR or metadata), it is a core feature of the skill.
- Ingestion points: User-supplied path evaluated in SKILL.md.
- Boundary markers: The agent is explicitly instructed to limit reports to 'one or two sentences' and 'advisory only'.
- Capability inventory: File system writes and local shell execution (sips, magick, swift) as defined in recipes.
- Sanitization: Shell scripts in the scripts/ directory utilize variable quoting to prevent command injection from file names.\n- [SAFE]: The skill uses native system utilities and standard development tools to perform legitimate image processing tasks. No evidence of data exfiltration, obfuscation, or persistence was found.
Audit Metadata