What problem does it solve?
Build and operate long-running background processes that start on boot, stay healthy, recover from crashes, and shut down gracefully across macOS, Linux, and Windows. This Skill addresses lifecycle management, process supervision, restart policies, logging and health endpoints, signal handling, and AI-specific concerns such as token-based LLM rate limiting and provider failover.
Core Features & Use Cases
- Cross-platform init guidance: When to use launchd vs systemd vs Windows services and how to place unit/plist files correctly.
- Resilience and restart policies: Recommendations for Restart, RestartSec, KeepAlive, throttling, and handling restart loops and zombie processes.
- Graceful shutdown and state persistence: Signal handling patterns, draining in-flight work, checkpointing rate-limit state, and safe resource cleanup.
- AI daemon patterns: Token-bucket rate limiting, per-provider circuit breakers, streaming handling, batching strategies, and saving token state on shutdown.
- Use Case: Run an AI inference daemon that persists token-bucket state, gracefully drains in-flight LLM requests on SIGTERM, fails over to secondary providers on 5xx, and avoids restart-thrashing in production.
Quick Start
Create a systemd unit with Type=notify and Restart=on-failure, implement a token-bucket per LLM provider with persistent state on disk, and add a SIGTERM handler that stops accepting new requests, drains in-flight work, flushes rate-limit state, and exits.
