api-design-reviewer
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from OpenAPI/Swagger specifications and incorporates content from these files, such as endpoint summaries and schema descriptions, into reports that the agent then processes. This creates a surface for indirect prompt injection where a malicious specification could attempt to influence agent behavior through the audit report.\n
- Ingestion points: The scripts
api_linter.py,api_scorecard.py, andbreaking_change_detector.pyall ingest user-provided JSON specification files.\n - Boundary markers: The generated text reports do not utilize explicit boundary markers or warnings to distinguish between tool-generated analysis and content extracted from the untrusted specification.\n
- Capability inventory: The scripts have file read and write capabilities as part of their standard execution flow. They do not perform network operations or direct shell execution.\n
- Sanitization: The scripts do not perform specific sanitization or escaping of the strings extracted from the API specifications before including them in the output.
Audit Metadata