beatra-ai-video-studio
Fail
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONCREDENTIALS_UNSAFEPERSISTENCEDATA_EXFILTRATION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill instructions direct the agent to download a ZIP archive from an external CDN (
cdn.beatra.ai) and execute the contained Python scripts (authorize.py,mcp_client.py). - [DYNAMIC_EXECUTION]: The software features a self-update mechanism that is on by default, which allows the client to automatically fetch and replace its own source files from a remote server without user review or approval during routine tool invocation.
- [CREDENTIALS_UNSAFE]: The authorization process involves obtaining and storing a persistent OAuth bearer token in
~/.beatra/credentials.json. This token has broad scopes, including the ability to spend prepaid credits (wallet:spend). - [COMMAND_EXECUTION]: The installation and verification steps require the agent to execute various shell commands, including
curlfor downloads,shasumfor integrity checks,unzipfor extraction, andcpfor file system modification. - [PERSISTENCE]: The skill establishes persistence by installing its scripts into the agent's local skills directory and creating a dedicated hidden directory (
~/.beatra/) to maintain state, installation references, and credentials across sessions. - [DATA_EXFILTRATION]: The skill's primary function involves uploading user-supplied files and prompts to external hosted infrastructure (
mcp.beatra.ai). It also transmits telemetry data, including the user's hostname and agent platform information, to the vendor's API.
Recommendations
- AI detected serious security threats
Audit Metadata