The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches guide content and metadata from Snowflake-Labs' official GitHub repositories (sfquickstarts and sfguide-* repositories) and official Snowflake domains.
[COMMAND_EXECUTION]: Executes SQL statements and Bash commands extracted from Quickstart guides. The skill is designed to run these commands specifically for the purpose of the tutorial, and it includes user confirmation steps (Learner mode) or summary reports (Builder mode).
[DATA_EXFILTRATION]: Conducts session diagnostics by querying standard Snowflake metadata (e.g., CURRENT_ACCOUNT_NAME, CURRENT_ROLE) to ensure the user's environment meets the Quickstart prerequisites. This data is used internally for diagnostics and is not sent to external third parties.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests and processes untrusted markdown content from external URLs to generate agent instructions.
Ingestion points: Fetches markdown and companion repository files from GitHub based on user-provided slugs (SKILL.md, Step 2 and Step 4).
Boundary markers: The skill does not explicitly define delimiters to isolate the guide content from the agent's core instructions.
Capability inventory: The skill has access to snowflake_sql_execute, Bash, and Write tools, which could be abused if malicious instructions were injected into the fetched content.
Sanitization: There is no evidence of sanitization or filtering of the fetched content beyond basic metadata parsing.
Assessment: Given that the sources are restricted to official Snowflake and Snowflake-Labs domains/repositories, this surface is considered a low risk inherent to the skill's primary function.

quickstart-guide