The Agent Skills Directory

[SAFE]: The skill consists of comprehensive architectural documentation and SQL templates intended to assist administrators in designing secure Snowflake environments. The content aligns with official security recommendations for data governance and role management.\n- [COMMAND_EXECUTION]: The skill generates SQL commands for account administration via the snowflake_sql_execute tool. Safety is ensured by explicit instructions requiring the agent to present all generated code for user review and wait for manual confirmation before proceeding with any changes.\n- [PROMPT_INJECTION]: No evidence of prompt injection or attempts to bypass AI safety guardrails was found. The skill includes human-in-the-loop checkpoints that effectively mitigate the risk of accidental or malicious instructions within user-provided requirements influencing administrative actions.\n- [CREDENTIALS_UNSAFE]: No hardcoded credentials, API keys, or secrets are present in the skill files. All SQL templates use generic placeholders (e.g., <DATABASE_NAME>) for environment-specific identifiers.\n- [EXTERNAL_DOWNLOADS]: The skill does not perform any network operations, external downloads, or remote data fetching. All instructional material is provided locally.\n- [REMOTE_CODE_EXECUTION]: Analysis revealed no remote code execution patterns, dynamic execution of untrusted scripts, or runtime compilation from external sources.

rbac