The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides examples for background jobs that ingest and process external data, creating an attack surface for indirect prompt injection if the processed data influences agent decision-making. \n
Ingestion points: ImportCSVJob in references/examples/examples.md (processes files via file_path) and SyncUserDataJob in references/examples/examples.md (processes data from ThirdPartyAPI). \n
Boundary markers: The simplified code examples do not include explicit boundary markers or instructions to ignore embedded commands in the processed data. \n
Capability inventory: The examples demonstrate capabilities for file system operations (File.delete), database writes, and external network requests across multiple scripts. \n
Sanitization: Input validation or sanitization logic for the external data is not explicitly demonstrated in the examples. \n- [DYNAMIC_EXECUTION]: The ReplayDLQJob example in references/advanced-features/error-handling.md demonstrates the use of .constantize to dynamically load job classes from database records (dlq_entry.job_class.constantize). While a standard Rails pattern, it requires strict access controls on the underlying data storage to prevent arbitrary class loading and execution.

rails-active-job