autogpt-agents

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

• Third-party content exposure detected (high risk: 0.75). Outsider free text can enter the LLM context via webhook-triggered graph execution: POST /api/v1/webhooks/{webhook_id} parses arbitrary JSON payload (data: "webhook payload") and then the graph’s LLM blocks consume node inputs derived from that payload.

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

• Potentially malicious external URL detected (high risk: 0.80). The quick-start and deployment instructions fetch and run external code (git clone https://github.com/Significant-Gravitas/AutoGPT.git and docker images like autogpt/platform-backend and autogpt/platform-frontend are pulled and executed via docker-compose), so remote content would be fetched and executed as part of setup/runtime.