public-sector-design-systems
Pass
Audited by Gen Agent Trust Hub on Aug 19, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill serves as an informational router, directing users to official government documentation and specific implementation skills. It does not contain any executable scripts (Python, Node.js, or Shell) and does not request access to restricted tools.
- [DATA_EXPOSURE_AND_EXFILTRATION]: All external references in the catalog target official government domains (e.g., .gov, .mil) or well-known, trusted source control platforms like GitHub (e.g., github.com/uswds, github.com/18F). No sensitive file paths or hardcoded credentials were detected.
- [PROMPT_INJECTION]: The instructions contain strong negative constraints, such as 'Do not invent component names' and 'Do not treat third-party catalogs... as statewide authority,' which improve the reliability of the agent and prevent hallucinations. No malicious override patterns were found.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests user input regarding jurisdictions and products to look up entries in its internal catalog. Since the skill lacks capabilities for file writing, network exfiltration, or command execution, the surface for indirect injection is not exploitable.
- [REMOTE_CODE_EXECUTION]: There are no patterns involving the download and execution of remote scripts (e.g., curl|bash). References to external 'skills' are pointers to other agent capabilities within the platform environment, not arbitrary remote code.
Audit Metadata