break
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill instructions require the agent to generate and save a new source file (a 'harness page' or scratch route) within the project's codebase. This generated code imports and executes existing project components to perform visual stress tests.
- [INDIRECT_PROMPT_INJECTION]: The skill purposefully ingests and renders 'hostile' data scenarios (e.g., extreme text lengths, specific character encodings, and RTL scripts) to test component resilience. This data is untrusted and enters the rendering pipeline without sanitization.
- Ingestion points: Scenario data defined in
scenarios.mdis fed as props into components. - Boundary markers: No specific delimiters or warnings are used to isolate these inputs from the rendering context.
- Capability inventory: The agent has the ability to write files to the local directory and import/execute project-specific code.
- Sanitization: None provided, as the intent is to test the component's raw handling of edge-case data.
- [COMMAND_EXECUTION]: The process of building the harness and viewing the result implies interaction with the local filesystem and the project's development server or build environment.
Audit Metadata