The Agent Skills Directory

[COMMAND_EXECUTION]: The skill references the use of the goldsky CLI for pipeline validation and data inspection (e.g., goldsky turbo validate and goldsky turbo inspect). These commands are standard tools provided by the vendor for developers.
[EXTERNAL_DOWNLOADS]: The fetch_abi function allows the pipeline to download ABI and IDL definitions from remote URLs at runtime. This is a core requirement for decoding blockchain event logs and instructions.
[DATA_EXFILTRATION]: The handler transform type enables sending pipeline records to external HTTP endpoints for enrichment. This involves outbound network communication with user-defined services.
[CREDENTIALS_UNSAFE]: The skill uses the secret_name parameter to reference credentials for database sinks and HTTP handlers. This practice promotes the use of the platform's secret management system instead of hardcoding sensitive information.
[REMOTE_CODE_EXECUTION]: The type: script transform allows the execution of custom TypeScript logic within a WASM-based sandbox. The environment is restricted, with no access to the network or external imports.
[PROMPT_INJECTION]: The skill processes untrusted data from blockchain sources (logs and transactions). It mitigates potential indirect injection risks by using structured decoding with ABIs and IDLs.
Ingestion points: Raw blockchain data sources such as base.logs and solana.transactions_with_instructions.
Boundary markers: The system uses structured schema definitions (ABIs/IDLs) which act as a validation layer for incoming data.
Capability inventory: The skill facilitates CLI execution, network fetching, and sandboxed scripting.
Sanitization: Decoding functions like _gs_log_decode enforce schema validation on the processed data.

turbo-transforms