refactor-with-ai
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [SAFE]: No malicious behavior, obfuscation, or attempts to bypass security controls were identified. The skill follows software engineering best practices for code improvement.\n- [COMMAND_EXECUTION]: The skill references standard shell commands for testing ('npm test'), type checking ('npx tsc'), and mutation testing ('npx stryker'). These commands are part of a routine development workflow and are used safely within the documented process.\n- [INDIRECT_PROMPT_INJECTION]: The skill defines a surface for processing user-provided code via interpolation into prompts.\n
- Ingestion points: In SKILL.md under Step 2 and Step 3, user code is pasted into '[paste code]' placeholders.\n
- Boundary markers: Prompts use markdown code blocks as delimiters.\n
- Capability inventory: Verification commands like 'npm' and 'npx' are mentioned for post-refactoring checks in SKILL.md.\n
- Sanitization: Not applicable as these are templates for manual or guided use; no automated processing of untrusted input is performed by the skill itself.\n- [NO_CODE]: The skill consists exclusively of instructional markdown content and does not include any standalone scripts or executable files.
Audit Metadata