find-overengineering
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data in the form of code diffs and subsystems.
- Ingestion points: Code diffs and subsystems provided to the agent for review (SKILL.md).
- Boundary markers: The skill lacks explicit delimiters or instructions to ignore embedded commands within the processed code.
- Capability inventory: The instructions authorize the agent to "Make small rollbacks right away," which involves autonomous file modification.
- Sanitization: There is no mention of sanitizing or validating the contents of the diffs before processing.
- Risk: A malicious actor could craft a code diff that uses comments or specific structures to trick the agent into identifying critical security guards or input validations as "overengineering," leading to their autonomous removal. The instruction to "Keep checks on input we don't control" provides a conceptual safeguard but does not eliminate the underlying injection surface.
Audit Metadata