use-case-qa
Pass
Audited by Gen Agent Trust Hub on Sep 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external inputs including 'accepted journeys,' 'test data,' 'SPEC,' and 'SLICES' to perform validation tasks. Because the agent executes these inputs via powerful interfaces (APIs, CLIs, browsers, and adapters), there is a potential for malicious data within these inputs to influence the agent's behavior. While the skill includes identity matching constraints, it lacks specific instructions for sanitizing the content of the data being processed.\n
- Ingestion points: Input files and project entities such as SPEC, SLICES, test data, and accepted journey definitions (SKILL.md).\n
- Boundary markers: Instructions mandate matching exact identities (goal-map, SPEC, SLICES, etc.) but do not define explicit delimiters or warnings to ignore instructions embedded within the data.\n
- Capability inventory: The skill utilizes a broad range of interfaces including browser, API, CLI, and custom adapters.\n
- Sanitization: The skill mentions validation of 'integrity metadata,' but lacks explicit sanitization or escaping protocols for interpolated data.
Audit Metadata