skills/akillness/jeo-skills/animato/Gen Agent Trust Hub

animato

Warn

Audited by Gen Agent Trust Hub on Jul 31, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The core functionality relies on sending Python scripts to an Animato server's /api/run and /api/chat endpoints for execution. This is a design-level Remote Code Execution surface. The skill attempts to sanitize these scripts using scripts/validate_bpy_script.py, but blocklist-based validation can often be bypassed by sophisticated code.
  • [DATA_EXFILTRATION]: When using --mode server, scripts/animato_agent.py transmits the user's LLM API keys (Gemini or OpenAI) to the ANIMATO_SERVER. If a user is tricked into pointing the agent at a malicious server, their credentials would be compromised.
  • [COMMAND_EXECUTION]: The scripts scripts/animato_agent.py and scripts/selftest.py utilize subprocess.run to execute local Python scripts and tests. The test script also utilizes dynamic module loading via import. This represents a capability to run shell commands and manage processes on the host system.
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection where a malicious motion request could result in the generation of Python code that performs unauthorized file access or network operations on the host, potentially evading the static gate checks.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 31, 2026, 03:30 AM
Security Audit — agent-trust-hub — animato