productlane-automation

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTIONNO_CODE
Full Analysis
  • EXTERNAL_DOWNLOADS (LOW): The skill directs the agent to connect to an external MCP server at https://rube.app/mcp. This endpoint is not on the trusted sources list and serves as a provider for tool logic. Severity is lowered as this is the primary intended purpose of the skill.
  • PROMPT_INJECTION (LOW): The workflow presents an indirect prompt injection surface (Category 8c). The agent executes tools based on schemas and plans fetched dynamically from the Rube service via RUBE_SEARCH_TOOLS. Ingestion points: tool schemas and recommended plans from RUBE_SEARCH_TOOLS (SKILL.md); Boundary markers: None specified; Capability inventory: RUBE_MULTI_EXECUTE_TOOL, RUBE_REMOTE_WORKBENCH; Sanitization: None described.
  • NO_CODE (SAFE): The skill does not include any Python or Node.js scripts, minimizing the risk of direct malicious code execution from the skill package itself.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:49 PM