mcp-management

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly configures and invokes MCP servers that fetch and browse the open web (e.g., references/configuration.md lists "brave-search" and "puppeteer" and references/gemini-cli-integration.md plus SKILL.md show examples like "Search…" and "Navigate to https://example.com" and using Playwright/Puppeteer), so the agent ingests untrusted public web content which can directly influence tool selection and subsequent actions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill's runtime config (.claude/.mcp.json) explicitly runs remote packages via npx (e.g., "npx -y @modelcontextprotocol/server-puppeteer" and "npx -y @modelcontextprotocol/server-memory"), which fetches and executes remote code that exposes tools/prompts (see https://github.com/modelcontextprotocol/servers), so external content is executed at runtime and can directly control agent behavior.