The Agent Skills Directory

[SAFE]: The skill serves as a library of reference implementations for asynchronous task processing. Analysis of the provided code snippets across TypeScript, Python, and Ruby shows adherence to secure coding practices.
[DATA_EXPOSURE_AND_EXFILTRATION]: Configuration for external services (Redis) is managed via environment variables (e.g., process.env.REDIS_URL, ENV['REDIS_URL']) rather than hardcoded secrets. Examples using hardcoded strings for connection URLs are limited to localhost environments, which is standard for documentation purposes.
[EXTERNAL_DOWNLOADS]: The skill references well-known, established libraries such as Bull, Celery, Sidekiq, and PyTorch. No suspicious or unversioned external dependencies from untrusted sources are identified.
[COMMAND_EXECUTION]: Code examples focus on application-level logic for job processing, data transformation, and model training. No arbitrary shell command execution or unsafe subprocess spawning patterns were detected.
[INDIRECT_PROMPT_INJECTION]: While the skill defines patterns for processing external data (e.g., email templates, ETL records), it does not include instructions that would allow untrusted data to influence agent behavior. As a pattern-based skill, it defines the structure for handling data rather than a specific vulnerable implementation.

loom-background-jobs