ai-test-generation
Pass
Audited by Gen Agent Trust Hub on May 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions suggest that the agent can initialize its environment using the command 'npx playwright init-agents --loop=claude' to facilitate interaction between the AI and the browser.- [EXTERNAL_DOWNLOADS]: The skill references and downloads dependencies from well-known public registries, specifically the '@playwright/mcp' package and the 'playwright' CLI tool.- [PROMPT_INJECTION]: The pipeline is designed to ingest and process potentially untrusted external data such as PRDs, feature specifications, and bug reports, which creates a surface for indirect prompt injection.\n
- Ingestion points: User-provided input files like PRDs, bug reports, and code diffs mentioned in Step 1 and Step 6.\n
- Boundary markers: The associated prompt library uses triple-dash delimiters to separate system instructions from user-provided content.\n
- Capability inventory: The skill enables the agent to generate executable test code and run CLI commands for testing purposes.\n
- Sanitization: Security is managed via a mandatory 'Step 7: Human Review' which includes a checklist to identify hallucinations, incorrect logic, and implementation details before any code is merged.
Audit Metadata