read-github

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

• Third-party content exposure detected (high risk: 1.00). This skill explicitly fetches and ingests public, user-generated GitHub repository documentation and arbitrary external URLs via the gitmcp.io MCP tools (see SKILL.md workflow and scripts/gitmcp.py's fetch-docs, search-docs, and fetch-url/call of fetch_generic_url_content), so untrusted third-party content can be read and influence agent actions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill spawns npx mcp-remote pointed at gitmcp.io (e.g. https://gitmcp.io/owner/repo) at runtime and depends on that MCP server’s returned tool definitions and documentation (and on npx fetching/executing the mcp-remote package), which can inject instructions or execute remote code that control the agent.