repository-threat-model
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFECREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill instructions direct the agent to specifically locate and document sensitive assets including API keys, database credentials, PII, and environment files (e.g., .env) as part of Step 1. While this is performed for the purpose of threat modeling, it encourages the agent to search for and process high-value secrets.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from codebase repositories (routes, models, configs, manifests) which acts as an ingestion point for potential adversarial instructions. The workflow lacks specified boundary markers or sanitization steps to prevent embedded instructions in the analyzed code from influencing the agent's logic. The agent maintains file-system read and write capabilities while processing this untrusted content.
Audit Metadata