clonedeps

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

• Third-party content exposure detected (high risk: 0.80). The skill's workflow (Step 2 and Step 4 in SKILL.md) has the orchestrator request remote repository URLs from the librarian and then explicitly git-clone/fetch public HTTPS GitHub/GitLab repositories into .slim/clonedeps/repos/, meaning the agent will ingest and read untrusted third‑party source code that can materially influence its decisions.

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

• Potentially malicious external URL detected (high risk: 0.90). This skill explicitly instructs the orchestrator to fetch/clone remote repositories at runtime (e.g., https://github.com/opencode-ai/opencode.git) so the cloned source is read by agents for inspection — meaning fetched external content can directly influence agent prompts/behavior.