frame-macos-notification
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill demonstrates an attack surface where untrusted data could influence agent actions during file authoring.
- Ingestion points: The agent processes user-supplied design briefs and instructions returned by the
prepare_open_design_briefMCP tool as defined in SKILL.md. - Boundary markers: The instructions lack specific delimiters or safety warnings to isolate the design-related instructions from potentially adversarial content contained within the tool output or user brief.
- Capability inventory: The agent is directed to use its own file-editing tools to author content and call
register_open_design_artifactto finalize the creation. - Sanitization: No sanitization, validation, or filtering of the instructions received from the user or the external MCP tool is implemented before the agent proceeds to author files.
Audit Metadata