deep-research

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

• Third-party content exposure detected (high risk: 0.90). The SKILL.md workflow (notably "Step 3: Information Gathering" and the "Browser-Based Research Crawling" / tool inventory sections) explicitly instructs the agent to run web searches and load/extract content from arbitrary public URLs using web-search tools, mcp__web-reader__webReader, agent-browser, Playwright, or curl/wget and then synthesize those sources into findings and recommendations, so untrusted third‑party content from the open web is ingested and can materially influence tool use and decisions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill contains a runtime install instruction to fetch remote skill code via "git clone https://github.com/mikeng-io/agent-skills" when dependencies are missing, which would retrieve and install executable skill files that can directly control agent prompts/behavior and is presented as a required dependency—so it meets the criteria for a risky external runtime dependency.