refactor
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute commands such as npm run test:coverage and npx stryker run. These are used to evaluate code quality and ensure that refactoring does not break existing functionality.
- [INDIRECT_PROMPT_INJECTION]: The skill analyzes and modifies existing codebase files, which could contain malicious instructions designed to influence the agent's behavior.
- Ingestion points: The agent reads source files using Read, Glob, and Grep tools during the code audit and test assessment phases.
- Boundary markers: There are no specific delimiters or instructional guards defined to isolate code content from the agent's control logic.
- Capability inventory: The skill has access to sensitive tools including Bash, Write, Edit, and TodoWrite, which could be exploited if an injection attack is successful.
- Sanitization: While the process includes linting and test verification, there is no specific logic to sanitize or filter out prompt injection attempts within comments or metadata of the ingested code.
Audit Metadata