melech-prune
Pass
Audited by Gen Agent Trust Hub on Aug 24, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local verification commands such as
npm test,pytest,tsc, andcargo testto ensure the codebase remains functional after pruning.\n- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it ingests untrusted code from the user's repository and uses it to make decisions about deletions.\n - Ingestion points: Reads local source files, Git diffs, and project structures during the audit phase (Step 2).\n
- Boundary markers: The instructions do not specify explicit delimiters or "ignore instructions" warnings when processing code content.\n
- Capability inventory: The skill has the ability to delete code blocks, remove files, and execute shell-based testing and build tools.\n
- Sanitization: There is no mention of escaping or filtering logic for the code symbols or comments analyzed.
Audit Metadata