multi-agent-orchestration

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.90). The skill explicitly supports per-agent MCPs that fetch public web/search results (see references/patterns.md Pattern 8 with mcps/researcher-mcps.json using a "server-brave-search") and the orchestrator (references/orchestrator-template.sh: get_mcp_config and run_agent/run_agent_bg) passes those MCP configs into claude --print, meaning agents ingest untrusted third‑party web content which can influence planner/researcher outputs and downstream pipeline decisions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). Flagged because the example MCP config mcps/researcher-mcps.json is used at runtime via --mcp-config and contains a command that runs "npx -y @modelcontextprotocol/server-brave-search", which will fetch and execute remote code (runtime external dependency).