agency-senior-project-manager
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a local shell script
./qa-playwright-capture.shto capture screenshots for quality assurance. It passes local environment arguments includinghttp://localhost:8000andpublic/qa-screenshots. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from a specification file which could potentially contain malicious instructions.
- Ingestion points: Reads site specification data from
ai/memory-bank/site-setup.md. - Boundary markers: The instructions do not define clear delimiters or escaping mechanisms to separate the project specifications from the agent's core instructions.
- Capability inventory: The skill has the ability to write files to the system (
ai/memory-bank/tasks/) and execute shell commands (./qa-playwright-capture.sh). - Sanitization: There is no evidence of input validation or sanitization for the content read from the specification file.
- [SAFE]: The skill references images from well-known services (Unsplash and Picsum Photos) for frontend development tasks, which is standard practice and poses no identified risk.
Audit Metadata