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 the webhook trigger path: POST /api/v1/webhooks/{webhook_id} accepts arbitrary JSON payload ("data": "webhook payload") which is then used as graph inputs and ultimately fed into LLM blocks during Block.execute().

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

• Potentially malicious external URL detected (high risk: 0.80). The quick-start installation instructs users to "git clone https://github.com/Significant-Gravitas/AutoGPT.git" and then build/run the project (docker compose up -d --build), which fetches and executes third‑party code from that repository at install/runtime, so this is a runtime external dependency that can control agent behavior.