cortex-code
Pass
Audited by Gen Agent Trust Hub on Apr 8, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill documentation references official installation scripts for the Cortex Code CLI (from ai.snowflake.com) and the uv package manager (from astral.sh). These resources originate from well-known technology providers and are documented as part of the standard installation workflow for the skill's prerequisites.
- [COMMAND_EXECUTION]: The skill utilizes the Python subprocess module to execute the cortex CLI tool and uv. These operations are essential for the skill's primary function of integrating with Snowflake's specialized agents. The execution logic is protected by a security wrapper that predicts tool requirements and implements user approval prompts.
- [DATA_EXFILTRATION]: A significant security feature in scripts/route_request.py proactively scans user prompts for sensitive file patterns, such as SSH keys, AWS/Snowflake credentials, and .env files. If any are detected, the skill immediately blocks the request to prevent accidental data exposure to the external agent.
- [PROMPT_INJECTION]: The skill implements a PromptSanitizer in security/prompt_sanitizer.py that detects and removes common prompt injection attempts (e.g., 'ignore previous instructions') and sanitizes personally identifiable information (PII) such as emails and credit card numbers before any data is sent to the external CLI.
Audit Metadata