copilot-sdk

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

• Third-party content exposure detected (medium risk: 0.30). The GitHub Copilot SDK runtime ingests outsider-authored free text via the required workflow’s session.sendAndWait({ prompt: input }) / prompt: "..." API, where any caller-supplied prompt becomes LLM-readable input.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill configures an MCP server URL that is contacted at runtime and can supply pre-built tools and model context (thus influencing prompts/agent behavior): https://api.githubcopilot.com/mcp/