cowart-open-canvas
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No security issues were detected. The skill is designed to manage a UI widget via an MCP (Model Context Protocol) tool.
- [COMMAND_EXECUTION]: The skill mentions a local script
scripts/start-canvas.sh, but it explicitly instructs the agent to avoid executing it for normal operations, designating it only as a local development fallback. - [INDIRECT_PROMPT_INJECTION]: The skill has a minimal attack surface as it primarily triggers a specific UI rendering tool based on direct user requests. Standard agent guardrails are sufficient for managing these interactions.
- [DATA_EXPOSURE]: The workflow involves passing a project directory path (
projectDir) to the rendering tool. This is standard behavior for development-oriented agents to identify the active workspace and does not constitute unauthorized data exposure.
Audit Metadata