apm-package-author

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

• Third-party content exposure detected (high risk: 0.80). This skill's workflow explicitly instructs using "apm install owner/repo/…" and states "APM paths ... resolve against the GitHub repo at install time" (SKILL.md and references/manifest-patterns.md), meaning it fetches arbitrary public GitHub content (user-generated) which can include MCP configurations that the install process merges into .vscode/mcp.json and thus can materially influence subsequent tool behavior.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill instructs running apm to fetch GitHub-hosted skill/agent content at runtime (e.g., "apm install thomast1906/github-copilot-agent-skills --runtime vscode"), which causes APM to pull remote .github/skills/.md files that directly control agent prompts/instructions—this is a runtime dependency on external content.