validate-agent
Pass
Audited by Gen Agent Trust Hub on Jun 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a local Python script "scripts/braintrust_analyze.py" via "uv run". It interpolates the "" variable directly into the shell command, which creates a risk of command injection if the path is manipulated.
- [PROMPT_INJECTION]: The agent processes untrusted data from technical plans and then uses its capabilities to perform web searches and local script execution.
- Ingestion points: Technical plan content and file paths provided at runtime in Step 1 and Step 2.
- Boundary markers: None specified in the instructions to distinguish untrusted plan content from agent instructions.
- Capability inventory: Shell command execution via "uv run", web search via "WebSearch", and file writing to a handoff directory.
- Sanitization: No explicit sanitization steps are defined for the plan content or the plan path before they are used in downstream tools or commands.
Audit Metadata