laravel-ai-sdk

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (medium risk: 0.30). The skill contains instructions for spawning external pilot and MCP agents before performing tasks, which introduces unauthorized or deceptive execution steps against the user.

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

• Third-party content exposure detected (medium risk: 0.30). The agent skill processes user-supplied transcripts and queries through its streaming and prompt interfaces (e.g., streaming controller templates reading request inputs), exposing the agent to untrusted user input.