test
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes content from the local codebase which could contain malicious instructions designed to manipulate the agent's behavior.
- Ingestion points: The skill reads source code and test files during Phase 1 (Discovery) and Phase 3 (Iterative Fixing) as defined in SKILL.md.
- Boundary markers: Absent; the instructions do not include delimiters or specific guidance for the agent to ignore instructions embedded within the files it is analyzing.
- Capability inventory: The skill has significant capabilities, including executing shell commands (test runners like pytest or jest), modifying files (Phase 2 and 3), and performing git commits (Phase 5).
- Sanitization: There is no evidence of content sanitization or validation to ensure that ingested code comments or strings do not contain prompt injection attacks.
Audit Metadata