The Agent Skills Directory

[COMMAND_EXECUTION]: The skill automates the use of the 'goldsky' CLI to manage the lifecycle of data pipelines and account authentication.
- Evidence: The builder workflow in SKILL.md utilizes commands such as goldsky project list, goldsky secret list, goldsky turbo validate, goldsky turbo apply, and goldsky turbo inspect to interact with the service.
[DYNAMIC_EXECUTION]: The skill involves generating configuration files that contain user-defined logic, including potential code transforms, which are subsequently executed by the service.
- Evidence: SKILL.md guides the agent to write YAML configurations (including SQL and TypeScript transforms) to disk and deploy them using the goldsky turbo apply command.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to translate high-level user goals into executable configurations, representing a managed attack surface.
- Ingestion points: The agent takes natural language descriptions of indexing goals (e.g., specific chains, contracts, and destinations) as input in SKILL.md.
- Boundary markers: The skill instructions provide clear demarcations for its scope, directing users to other specialized skills for diagnostics (/turbo-doctor) or syntax reference (/turbo-pipelines).
- Capability inventory: The skill can write files to the local file system and execute CLI commands that deploy infrastructure changes.
- Sanitization: The workflow explicitly mandates a validation check via goldsky turbo validate to ensure the integrity and correctness of the generated configuration before it is presented or applied.

turbo-builder