api-design-reviewer
Warn
Audited by Gen Agent Trust Hub on Apr 1, 2026
Risk Level: MEDIUMPROMPT_INJECTIONSAFE
Full Analysis
- [PROMPT_INJECTION]: The skill body in SKILL.md identifies the maintainer as 'Claude Skills Team', which is misleading given the author context 'borghei'. This impersonation may cause users or agents to misjudge the source and reliability of the skill.
- [PROMPT_INJECTION]: The static analysis tools ingest untrusted OpenAPI JSON data, representing an indirect prompt injection surface if the agent interprets malicious content within the analyzed files.
- Ingestion points: api_linter.py, api_scorecard.py, and breaking_change_detector.py read external JSON files.
- Boundary markers: Absent; no isolation of untrusted data from tool output is implemented.
- Capability inventory: Scripts have file-write access via the --output parameter.
- Sanitization: Data is parsed using standard JSON libraries without specific filtering for embedded instructions.
- [SAFE]: The provided scripts for API design analysis are implemented using standard Python libraries and do not contain remote code execution, unauthorized data exfiltration, or malicious persistence mechanisms.
Audit Metadata