hyper-cleanup
Pass
Audited by Gen Agent Trust Hub on Nov 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes local shell commands to execute the project's existing test suite for verification. This is a core functionality required to establish a 'behavior lock' before any code is modified and to verify correctness after cleanup. The instructions strictly limit this to the project's own test or check commands.- [DYNAMIC_EXECUTION]: The bundled script
mutation-proof.shperforms controlled, in-place text mutations on source files (usingawk) to verify that tests are capable of detecting regressions. This process is secured by mandatory SHA-256 checksum verification, ensuring that every modified file is restored byte-for-byte to its original state.- [INDIRECT_PROMPT_INJECTION]: As the skill processes untrusted data from a codebase, it includes explicit guardrails to treat external content (such as code comments or issue text) strictly as evidence for cleanup, rather than as instructions that could widen the skill's authority or bypass safety rules. It also defines adversarial evaluation cases to ensure compliance with this boundary.
Audit Metadata