The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes job data from external sources that can influence agent or system behavior.
Ingestion points: Job data is ingested through the JobService methods and the /api/users/route.ts endpoint, which takes JSON input from HTTP requests.
Boundary markers: No explicit delimiters or instructions are provided to the workers to ignore potentially malicious instructions embedded in job data (e.g., in email templates or webhook payloads).
Capability inventory: The workers have access to network operations (fetch in webhook.worker.ts), file system processing via sharp in image.worker.ts, and external storage via S3Client in image.worker.ts.
Sanitization: The provided code snippets do not include sanitization or validation logic for the template, variables, or url fields before they are used in sensitive operations.
[DATA_EXFILTRATION]: The webhookWorker executes network requests to arbitrary URLs defined in the job data. While this is the core purpose of a webhook processor, it could be leveraged to exfiltrate information if the job creation process is not strictly controlled.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of several well-known and trusted Node.js packages including bullmq, ioredis, sharp, and the AWS SDK.

queue-job-processor