code-refactoring
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied source code to analyze behavior and perform refactorings. This creates a surface where malicious instructions could be embedded in code comments or data structures to influence agent behavior during analysis.\n
- Ingestion points: User-provided code analyzed in Step A: Understand Current Behavior (SKILL.md).\n
- Boundary markers: The skill uses logical headers for analysis but lacks explicit instructions to ignore embedded prompts in the source.\n
- Capability inventory: The skill triggers shell commands including npm, npx, and codex-cli to validate code changes (SKILL.md).\n
- Sanitization: No sanitization of the input code is described.\n- [COMMAND_EXECUTION]: The skill instructs the agent to execute various shell commands for testing, type-checking, and linting the refactored code.\n
- Evidence: Instructions to run npm test, npx tsc, npm run lint, and specialized CLIs like codex-cli and ask-gemini (SKILL.md).
Audit Metadata