The Agent Skills Directory

[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface through its code generation templates.\n
Ingestion points: Templates in the templates/ folder (e.g., task.template.py, beat_schedule.template.py) ingest user-defined task names, parameters, and descriptions via Jinja-style placeholders.\n
Boundary markers: Absent. The templates do not include delimiters or instructions for the agent to ignore embedded commands in user-provided fields during code generation.\n
Capability inventory: The generated code defines Celery tasks and configurations which are executed by background workers with access to the application's environment, message brokers, and internal resources.\n
Sanitization: Absent. There is no evidence of input validation or escaping before interpolation into the templates.\n- [DATA_EXFILTRATION]: The file examples/fastapi_celery.example.py implements a webhook callback pattern in batch_process_task. This allows sending task results to a user-provided webhook_url, which establishes a potential SSRF or data exfiltration surface if the destination is not validated against a trusted whitelist.\n- [EXTERNAL_DOWNLOADS]: The skill documentation and examples recommend installing well-known, standard libraries including celery, fastapi, uvicorn, pydantic-settings, and httpx from official package registries.

using-celery