The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill reads the ~/.claude/settings.json configuration file to check for existing Snowflake MCP servers. This file is a sensitive configuration store for the agent platform.
[COMMAND_EXECUTION]: On Windows systems, the skill executes PowerShell scripts using the -ExecutionPolicy Bypass flag to allow installation scripts to run, which overrides standard system security settings.
[COMMAND_EXECUTION]: The skill searches the local filesystem using find or Get-ChildItem and then executes the path returned by these commands. This represents a risk of executing unauthorized code if a malicious file is placed in a matching path.
[EXTERNAL_DOWNLOADS]: Clones the snowflake-ai-kit repository from the official Snowflake-Labs GitHub organization as part of the setup process.
[REMOTE_CODE_EXECUTION]: Executes an installation script (install.sh or install.ps1) provided by the Snowflake-Labs repository.
[PROMPT_INJECTION]: The skill exhibits a vulnerability surface for indirect injection by executing files found via filesystem search without sanitization.
Ingestion points: Results from find and Get-ChildItem commands (SKILL.md).
Boundary markers: None identified.
Capability inventory: Execution of arbitrary shell and PowerShell scripts (SKILL.md).
Sanitization: No validation or verification of the found script's integrity or origin before execution.

cortex-setup