github-agentic-workflows

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

• Third-party content exposure detected (high risk: 0.85). Outsider free text can enter the LLM context via GitHub event payloads such as on: issue / on: issue_comment (e.g., issue bodies and comments authored by others) that the workflow/agent reads to generate outputs.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly states the repository uses the GitHub MCP Insiders endpoint (https://api.githubcopilot.com/mcp/insiders) configured in .github/copilot-mcp.json, which is invoked at runtime by the Copilot/MCP runtime to obtain model responses and therefore directly controls agent prompts/execution as a required external dependency.