prime-agent-docs

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

• Third-party content exposure detected (medium risk: 0.30). In ACP mode, an outsider can send arbitrary session/prompt JSON-RPC messages (free-text prompts) over stdin/stdout that Prime Agent will include as user prompt content for the model during turn execution.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill documentation contains explicit installer and development commands that fetch and execute remote code at runtime — e.g., curl -fsSL https://app.primeintellect.ai/prime-agent/install.sh | sh and git clone https://github.com/PrimeIntellect-ai/prime-agent followed by npm ci and running the checked-out script — which are high-confidence runtime fetch-and-execute vectors.