bep-best-openapi-design

Pass

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute a bundled Python script (scripts/check_resource_paths.py) to validate API paths.
  • Evidence: The script uses only Python standard libraries (argparse, re, sys, pathlib, urllib.parse) and performs local path analysis without network or sensitive file access.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data like OpenAPI descriptions and source code to generate design recommendations or implementation code.
  • Ingestion points: Reads repository instructions, existing routes, OpenAPI descriptions, schemas, clients, tests, and API conventions to inform its design decisions.
  • Boundary markers: Not explicitly defined for external content, though the skill enforces a strict "fail closed" policy for incomplete resource models.
  • Capability inventory: The skill generates OpenAPI contracts, implementation code, and executes a local path-checking script.
  • Sanitization: No specific sanitization or filtering of external content is detailed before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 3, 2026, 09:03 PM
Security Audit — agent-trust-hub — bep-best-openapi-design