threat-modeling

Pass

Audited by Gen Agent Trust Hub on Sep 18, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data, specifically source code, architecture diagrams, and business context, to perform security analysis.
  • Ingestion points: System architecture extraction (SKILL.md Step 2) and repository-wide code analysis involve reading arbitrary files from a user's environment.
  • Boundary markers: The skill instructions do not specify the use of delimiters or 'ignore embedded instructions' prompts when interpolating code content into the LLM context.
  • Capability inventory: The skill possesses the capability to make network requests to local and vendor API endpoints and execute MCP tools (e.g., gitnexus) to fetch data or post assessments.
  • Sanitization: While the skill includes a 'PhoenixDataShieldingService' to protect proprietary intellectual property from exposure, it lacks documented sanitization or escaping mechanisms to prevent instructions within the analyzed code from overriding the agent's primary directives.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 18, 2026, 03:58 AM
Security Audit — agent-trust-hub — threat-modeling