diagnosing-react-code
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes
npx -y react-doctor@0.9.17to perform code analysis. This is a standard method for running Node.js utilities. - [EXTERNAL_DOWNLOADS]: The skill downloads the
react-doctorpackage from the npm registry. The instructions mandate using a pinned version (0.9.17) to ensure consistency and prevent unexpected behavior from updates. - [DATA_EXPOSURE]: The documentation notes that
react-doctorcommunicates with an external score API and Socket.dev for supply-chain analysis. It correctly identifies this behavior and provides the--no-scoreand--no-supply-chainflags as options for users who wish to keep all data local. - [INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface by reading and processing local code files and tool diagnostics. The agent is instructed to fix code based on these outputs.
- Ingestion points: Project files (
package.json, React components, hooks) and the output of thereact-doctorcommand. - Boundary markers: Not explicitly defined for the tool output, though the agent is given structured instructions on how to interpret the results.
- Capability inventory: File-write access (to fix diagnostics) and command execution (to run the scanner).
- Sanitization: The skill relies on the agent's internal logic to filter and apply fixes to the code based on the diagnostic rules.
Audit Metadata