flowstudio-power-automate-build

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The Python helper code in SKILL.md initiates network requests to an external service endpoint that is not on the whitelisted domains list.
  • Evidence: MCP_URL = "https://mcp.flowstudio.app/mcp" in SKILL.md.
  • Evidence: Usage of urllib.request.urlopen to send flow definitions and configuration data to this endpoint.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the MCP server (such as flow definitions and connection details) and uses it to construct subsequent tool calls that modify the environment.
  • Ingestion points: list_live_flows, get_live_flow, and list_live_connections in SKILL.md retrieve data from the external server.
  • Boundary markers: The skill includes instructions to "Always call list_skills / tool_search first" to verify available tool names and schemas before proceeding.
  • Capability inventory: The skill utilizes update_live_flow, trigger_live_flow, and set_live_flow_state to perform modifications and executions in the Power Automate environment.
  • Sanitization: No explicit sanitization or filtering logic is provided for the data retrieved from the API before it is processed by the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 07:26 AM