autogpt-agents

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches source code from the official AutoGPT GitHub repository.\n
  • Evidence: git clone https://github.com/Significant-Gravitas/AutoGPT.git in SKILL.md.\n- [COMMAND_EXECUTION]: Uses Docker, npm, and poetry for platform deployment and environment management.\n
  • Evidence: Usage of docker compose, npm install, and poetry run in SKILL.md and troubleshooting.md.\n- [PRIVILEGE_ESCALATION]: Documentation includes maintenance commands requiring administrative access for system services and firewalls.\n
  • Evidence: sudo systemctl start docker and sudo ufw allow 8001 in troubleshooting.md.\n- [DYNAMIC_EXECUTION]: Allows registration of custom functional blocks and dynamic construction of agent graphs.\n
  • Evidence: Examples for MyCustomBlock and create_dynamic_graph in advanced-usage.md.\n- [INDIRECT_PROMPT_INJECTION]: The platform architecture includes an attack surface for instructions embedded in processed data.\n
  • Ingestion points: Data entering through webhooks and integration connectors in SKILL.md.\n
  • Boundary markers: Block implementation examples lack explicit delimiters for external data interpolation.\n
  • Capability inventory: Support for custom block logic, network requests, and database access across advanced-usage.md and troubleshooting.md.\n
  • Sanitization: Mention of input sanitization for monitoring purposes, but lacks universal validation logic for agent execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 07:07 PM
Security Audit — agent-trust-hub — autogpt-agents