agentic-ecology-colab-cli
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill requires the installation of
google-colab-cliviauvorpip. This tool is used as the primary interface to provision and manage remote compute resources. - [COMMAND_EXECUTION]: The technical reference documentation includes shell commands for performing Google Cloud authentication (
gcloud auth) and using thecolabCLI for session lifecycle management, including starting, stopping, and mounting storage. - [DYNAMIC_EXECUTION]: The skill utilizes
colab execto transmit and execute local Python scripts onto a remote Google Colab virtual machine. This is the core functionality intended for heavy workload offloading. - [INDIRECT_PROMPT_INJECTION]: The skill workflow monitors remote session logs for authentication URLs to facilitate the OAuth flow. Ingesting untrusted output from remote logs into the agent's context establishes an attack surface where malicious strings in logs could attempt to influence the agent.
- Ingestion points: Monitoring logs for generated OAuth URLs during authentication and Drive mounting (SKILL.md, steps 2 and 4).
- Boundary markers: Absent; the instructions do not specify delimiters for log content.
- Capability inventory: Includes remote command execution (
colab exec), package installation (colab install), and filesystem access via Google Drive (colab drivemount). - Sanitization: None mentioned; the workflow relies on the user to manually verify and provide authorization codes.
Audit Metadata