The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides patterns for building workers that process event data from external sources (Redis/Kafka). This creates a standard surface for indirect prompt injection where malicious instructions could be embedded in the processed data payloads.
Ingestion points: services/cio-queue-worker.ts and lib/scaled-processor.ts process data objects containing user IDs and event attributes.
Boundary markers: None explicitly implemented in the code snippets, though the skill mandates testing in isolated staging environments.
Capability inventory: Includes network operations via the Customer.io API and shell execution for operational tasks (npm, kubectl).
Sanitization: Standard JSON parsing is employed without additional content-level sanitization for LLM instructions.
[SAFE]: Secret management follows industry best practices by utilizing environment variables (process.env.CUSTOMERIO_TRACK_API_KEY, etc.) rather than hardcoding credentials.
[SAFE]: Dependencies such as bottleneck, bullmq, and kafkajs are well-known, reputable packages for the intended functionality.
[SAFE]: The skill includes explicit operational safety guidance, specifically advising users to never perform load tests against production environments or live recipient lists.

customerio-load-scale