fuck-over-engineering
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill identifies over-engineered patterns while explicitly instructing the agent to preserve security-critical logic, such as trust-boundary validation and accessibility controls.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests codebase files as untrusted data. It mitigates this risk by requiring human-in-the-loop confirmation before modifying any source files, as specified in the output instructions.
- [COMMAND_EXECUTION]: The skill utilizes standard file system reading and documentation search tools. It does not attempt to execute arbitrary shell commands or acquire elevated privileges.
Audit Metadata