patent-draft-agent

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

• Third-party content exposure detected (low risk: 0.10). The workflow requires patent and web search via google-patents-mcp and Exa (or web browsing) to ingest search results and crawled pages, which represent outsider-authored web content fetched via search queries.

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

• Potentially malicious external URL detected (high risk: 0.70). The skill executes npx via an MCP configuration file and script to install and run third-party npm packages (@kunihiros/google-patents-mcp and exa-mcp-server) at runtime, constituting a weak/unknown/suspicious operational dependency.