browse-flows

Pass

Audited by Gen Agent Trust Hub on Oct 27, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [Indirect Prompt Injection Surface]: The skill retrieves and processes external data from Power Automate environments, which represents a potential surface for indirect prompt injection.
  • Ingestion points: Data enters the context via mcp__flowagent__list_environments, mcp__flowagent__list_flows, and flow detail tools in SKILL.md.
  • Boundary markers: There are no explicit delimiters or specific instructions provided to the agent to disregard instructions embedded within flow names or descriptions.
  • Capability inventory: The skill has access to shell tools (Bash) and file system operations (Read, Glob, Grep), alongside specialized Power Automate management tools.
  • Sanitization: The instructions do not specify any sanitization or validation of the metadata retrieved from the Power Automate API.
  • Context: This is a common design pattern for tools that manage external resources. The risk is minimized by the agent's internal safety guardrails and the fact that the data is primarily used for presentation in tabular format.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 27, 2026, 09:12 PM