learn

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

• Third-party content exposure detected (high risk: 0.90). The SKILL.md /learn flow explicitly clones arbitrary URLs (Step 0: "ghq get [url]") and then launches agents to "Explore [REPO_PATH]" so the agent will ingest and act on untrusted, public repository contents, enabling indirect prompt injection.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill runs "ghq get [url]" at runtime to clone a user-supplied git repository and then has its agents read/analyze the repository contents, so external repo content fetched during runtime is injected into the agent context and can directly control prompts.