react-modernization
Warn
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: MEDIUMPROMPT_INJECTION
Full Analysis
- [Indirect Prompt Injection] (MEDIUM): The skill is designed to process and refactor React codebases which are untrusted data sources. This creates a surface where malicious instructions embedded in code comments or metadata could influence the agent's behavior. 1. Ingestion points: React source code files (components, hooks) and configuration files. 2. Boundary markers: Absent; no instructions are provided to separate user code from agent instructions. 3. Capability inventory: Refactoring tasks typically require file-system write access and execution of codemod CLI tools. 4. Sanitization: Absent; no measures are defined to filter or ignore natural language instructions found within the source files.
Audit Metadata