advai-skill

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

• Third-party content exposure detected (high risk: 0.75). The required workflow can install/sync a skill from a user-supplied GitHub URL (advai skill install <github-url> / advai skill sync <name>), which at runtime will fetch and ingest outsider-authored repository files (e.g., skill definition text) into the agent’s LLM context via the target platform’s sync mechanism.

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

• Potentially malicious external URL detected (high risk: 0.90). This skill invokes runtime installs from GitHub (e.g., advai skill install such as https://github.com/example/skills-repo) and requires the advai CLI from https://github.com/Advai-X/advai-cli, which means remote repository content fetched at runtime can contain skill definitions/prompts that directly control agent behavior.