The Agent Skills Directory

[COMMAND_EXECUTION] (SAFE): The snowflake-deploy companion agent executes standard Snowflake CLI commands to automate Native App deployment workflows. This is a functional requirement for the skill's specialist agent role.
[CREDENTIALS_UNSAFE] (SAFE): The skill provides detailed instructions on Snowflake JWT and RSA key-pair authentication. It correctly uses placeholders for sensitive information and provides proactive warnings against sharing credentials in logs or configuration files.
[DATA_EXFILTRATION] (SAFE): Analysis of the REST API guide and Cloudflare Workers templates shows no evidence of data exfiltration. All network operations are directed towards official Snowflake endpoints or user-defined infrastructure.
[INDIRECT_PROMPT_INJECTION] (SAFE): The deployment agent ingests data from local project configuration files (snowflake.yml, manifest.yml). While this presents a theoretical indirect prompt injection surface, the risk is mitigated by the agent's limited scope and the use of the official snow CLI for processing.
Ingestion points: snowflake.yml, app/manifest.yml, app/setup_script.sql (referenced in agents/snowflake-deploy.md)
Boundary markers: Absent
Capability inventory: Bash tool usage for CLI operations and file inspection
Sanitization: Absent

snowflake-platform