agentic-research-orchestration

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

• Third-party content exposure detected (high risk: 0.90). The skill's Phase 4 "URL Relay Protocol" (SKILL.md / agentic-research-orchestration.md and references/agent-spawn-patterns.md and the orchestration transcript) explicitly ingests and relays open-web URLs discovered by ExaAI and Firecrawl (e.g., blogs, GitHub, StackOverflow, documentation sites) to NotebookLM and triggers crawls, meaning the agent reads and acts on untrusted third‑party content that can change tool behavior.

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

• Potentially malicious external URL detected (high risk: 0.80). The orchestrator explicitly relays and instructs fetching of external sites at runtime (e.g., https://github.com/langchain-ai/langgraph), and those fetched pages/repos are added into the NotebookLM notebook and/or auto-embedded by Firecrawl—meaning remote content is ingested into the model context and can directly influence prompts/analysis.