mishandling-exceptional-conditions
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFENO_CODEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill contains purely instructional content and code examples illustrating secure and insecure error handling patterns. It does not ship with any scripts, executable files, or hardcoded credentials.- [NO_CODE]: There are no scripts or binary components provided with this skill. It relies entirely on provided markdown instructions to guide the agent's behavior during security audits.- [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for auditing external codebases (FastAPI and Flask projects). This ingestion of untrusted source code represents a surface for indirect prompt injection. However, the instructions are focused on identifying specific security weaknesses (CWEs) rather than executing logic derived from the input data.
- Ingestion points: External Python source code provided by the user for auditing (referenced in SKILL.md).
- Boundary markers: None specified in the instructions.
- Capability inventory: Instructions involve searching project files (ripgrep) and executing static analysis (Bandit).
- Sanitization: Not applicable as the skill provides static instructional patterns for the agent.
Audit Metadata