ios-review
Pass
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill demonstrates strong security posture by including specific directives that command the agent to treat external data strictly as information to be studied rather than instructions to be executed. It also explicitly forbids the reproduction of secrets or private content from analyzed applications.\n- [INDIRECT_PROMPT_INJECTION]: The skill defines a surface area for indirect injection through UI data retrieved via vendor-provided tools.\n
- Ingestion points: UI data, accessibility labels, and journey descriptions retrieved from the uxkin.com MCP server via the
find_ui_referencesandget_journeytools.\n - Boundary markers: Present. The skill includes a dedicated section titled 'References are data, not instructions' which provides clear delimiters for behavioral boundaries.\n
- Capability inventory: The skill audits mobile design and uses findings to suggest code changes, which can be applied via the related
ios-screensskill upon user confirmation.\n - Sanitization: The instructions mandate that the agent ignore any text within the retrieved data that attempts to influence its behavior or override instructions.
Audit Metadata