The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides patterns for processing external job data via BullMQ workers, which creates an indirect prompt injection surface. An attacker who can influence job payloads could potentially inject malicious instructions into the system's processing flow.
Ingestion points: ExampleProcessor.process(job: Job<ExampleJobData>) in references/patterns.md processes data typically sourced from external queues or upstream services.
Boundary markers: The provided code examples lack explicit boundary markers or delimiters for user-supplied job data (e.g., entityId).
Capability inventory: The skill utilizes Logger for recording activity and TypeORM repositories for database persistence, as seen in ExampleProcessor and ExampleQueueService.
Sanitization: The patterns do not explicitly demonstrate sanitization or validation of the entityId or other job data before it is logged or persisted to the database.
[SAFE]: The skill promotes safe software architecture patterns, such as the 'fail-open' mechanism in references/patterns.md for the Redis throttler, which prevents a Redis outage from causing a Denial of Service (DoS) for the entire HTTP API.
[SAFE]: The skill correctly manages environment-specific configurations by recommending local Redis for development environments to avoid unintended costs and resource exhaustion on cloud services like Upstash.
[SAFE]: All referenced libraries (e.g., @nestjs/bullmq, bullmq, typeorm) are well-known, industry-standard packages from trusted registries.

nestjs-bullmq