advai-cli

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

• Third-party content exposure detected (high risk: 0.70). The skill’s runtime path includes installing and inspecting external CLIs from a user-supplied GitHub URL (advai cli install <github-url> / advai cli info <name>), which can cause the agent to ingest free-form text from outsider-authored repository content (e.g., README/metadata) into the LLM context via the advai CLI’s outputs.

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

• Potentially malicious external URL detected (high risk: 0.90). This skill invokes advai cli install at runtime and explicitly requires/installing advai from https://github.com/Advai-X/advai-cli (and example installs like https://github.com/example/tooling-repo), which fetches and installs external GitHub repositories that can contain and execute remote code.