The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill clones configuration and setup scripts from the Snowflake-Labs official GitHub repository (github.com/Snowflake-Labs/sf-solutions). This operation targets a verified vendor repository and is documented neutrally as a core functional requirement.
[COMMAND_EXECUTION]: Uses system tools to perform repository cloning (git clone) and executes SQL setup scripts against the user's Snowflake account using the snowflake_sql_execute tool. The skill enforces a mandatory stopping point for user review before any execution occurs.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data which serves as a potential surface for indirect instructions.
Ingestion points: Metadata and instructions are ingested from manifest.json and README.md files within the cloned repository.
Boundary markers: The skill architecture includes an explicit plan presentation step that requires the user to manually verify and approve the installation details.
Capability inventory: The skill has permissions to execute SQL, search objects, and perform basic file system and network operations.
Sanitization: The implementation relies on human-in-the-loop verification to validate the intended actions before they are performed.

solutions-installer