MCP Development Skill

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

• Third-party content exposure detected (high risk: 0.90). The SKILL.md explicitly shows the agent reading and acting on third-party, user-generated sources (e.g., resource URIs like "github://repo/owner/repo/readme", a "search_issues" tool, and the "Common MCP Servers" list that includes "github" and "puppeteer" for web scraping), meaning the agent will ingest untrusted public web/social content as part of its workflow and that content can influence tool calls and decisions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill includes a runtime command that fetches and executes remote code — the Claude Desktop example uses "npx -y @modelcontextprotocol/server-github", which would download and run a third‑party MCP server that can provide tools/prompts to the agent and thus directly control agent instructions.