swarm-advanced

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

• Third-party content exposure detected (high risk: 0.75). The skill’s runtime workflow includes “web-search” / “content-extraction” style research tasks (in references/patterns.md under Research Swarm Phase 1), which implies fetching and extracting public web content at runtime and feeding that extracted free text into the agents’ LLM context (indirect prompt injection risk).

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

• Potentially malicious external URL detected (high risk: 0.80). The skill's Quick Start explicitly runs remote package code (npm install / npx claude-flow@alpha -> npx claude-flow@alpha mcp start) which will fetch and execute the external claude-flow package relied on at runtime (repository: https://github.com/ruvnet/claude-flow), so this is a runtime dependency that executes remote code.