langgraph-orchestration

Pass

Audited by Gen Agent Trust Hub on Apr 7, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides patterns for running Model Context Protocol (MCP) servers using npx. This is a standard and documented method for initializing MCP servers within the ecosystem.
  • [EXTERNAL_DOWNLOADS]: The skill references several well-known libraries and frameworks for agent orchestration, including langgraph, langchain, and @modelcontextprotocol/server-filesystem. These are established tools within the AI development community.
  • [PROMPT_INJECTION]: The skill describes building agents that ingest data from external sources such as the internet, Jira, and GitHub, which creates a potential surface for indirect prompt injection. The skill mitigates this risk by providing explicit instructions and patterns for implementing human-in-the-loop (HITL) approval gates (interrupt()) and strict tool isolation boundaries between agents.
  • Ingestion points: references/mcp-bridging.md (Jira/GitHub), references/deep-agents.md (internet search/fetch URL).
  • Boundary markers: Recommended usage of interrupt() and approval_gate in references/interrupts-hitl.md.
  • Capability inventory: Filesystem access, command execution (npx), and external network requests.
  • Sanitization: The skill promotes narrow tool scopes for specialist workers and manual approval for risky side effects.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 7, 2026, 09:03 PM