api-reviewer
Pass
Audited by Gen Agent Trust Hub on Mar 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: Employs the
Globtool to search for PRDs and existing contracts within the filesystem (Phase 0.2). Additionally, it suggests execution of local linting and automated checks (Phase 0.4). - [PROMPT_INJECTION]: Establishes an indirect prompt injection surface by ingesting and analyzing untrusted external content from API contracts and PRD documents.
- Ingestion points: Content extracted from PRDs, boundary confirmations, and API specifications (Phase 0 and 1).
- Boundary markers: None identified. The skill lacks instructions to isolate document content from the agent's logic.
- Capability inventory: File system traversal (
Glob), document reading, and interactive questioning. - Sanitization: No input validation or filtering of document text is specified before processing.
Audit Metadata