arete-consume

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

• Third-party content exposure detected (high risk: 0.80). The skill's workflow (SKILL.md steps 2, 5, and "Generating SDK types for custom stacks" plus references/typescript-api.md) has the agent discover and connect to arbitrary Arete stacks (including via external wss:// URLs), ingesting live, potentially public or user-provided entity data over WebSockets which the agent is explicitly instructed to read and use to drive code, streaming logic, and decision-making — exposing it to untrusted third-party content that could carry indirect prompt-injection vectors.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill's prerequisites instruct running remote installers that are fetched and executed (curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y --no-modify-path and downloading https://win.rustup.rs/x86_64 then executing it) to install the Rust toolchain/CLI which the skill requires, so these URLs execute remote code during setup.