install-rules

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly supports fetching/cloning a user-specified Git repo and parsing its INDEX.md or listing *.md files to build rules (see SKILL.md "Specified Git repo" / "Resolve source" steps), which are arbitrary third‑party/user-generated contents the agent must read and that can directly alter installed rules and agent behavior.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly fetches user-specified Git repositories at runtime (e.g., https://github.com/owner/repo.git or git@github.com:owner/repo.git) to read rule Markdown that are installed as agent “rules” (passive constraints) which directly control agent behavior/prompts, so remote repo content can control prompts during runtime.