broken-access-control

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides instructions for using well-known security CLI tools such as Bandit, OWASP ZAP, and TruffleHog to audit local application code and environments. These commands are standard for the stated purpose of security testing.
  • Evidence found in SKILL.md: bandit -r . -t B105,B106,B107,B501,B601, zap-baseline.py -t http://localhost:8000, and trufflehog git file://. --only-verified.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze user-provided source code (FastAPI/Flask) to identify security vulnerabilities. This presents a potential attack surface if the code being analyzed contains malicious instructions targeting the agent, although this is inherent to the auditing use case.
  • Ingestion points: The agent processes code provided by the user in response to audit requests (e.g., 'audit my routes').
  • Boundary markers: None explicitly defined in the provided snippets to delimit user code from agent instructions.
  • Capability inventory: The skill uses logic to print routes, check dependencies, and perform object-level permission checks on the provided code.
  • Sanitization: No specific input sanitization for user code is described in the audit workflow.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 08:14 PM
Security Audit — agent-trust-hub — broken-access-control