What problem does it solve? Navigating lci's daemon architecture is hard: the per-root index server involves auto-spawn logic, a Unix-socket HTTP route table, an instance registry, and multiple shutdown paths (idle timeout, RSS cap, root-gone reaper, LRU eviction) spread across many source files. This Skill maps every surface to exact file and line locations so you can trace, debug, or modify server behavior without grepping the whole codebase. ## Core Features & Use Cases - Complete code map: Tables linking every CLI verb (lci server, servers, shutdown, status), socket path helper, registry function, and HTTP route to precise source locations in src/cli/ and src/server/. - Config and invariants reference: Documents server { idle_timeout_sec, max_instances, max_rss_mb } defaults plus traps like lock-free read paths, /ping not stamping activity, and exactly-once shutdown semantics. - Probe recipes and tests: Ready-to-run commands for listing servers, curling the Unix socket directly, verifying loopback-only posture, and running targeted gtest filters. - Use Case: You need to debug why an lci server did not exit after its project root was deleted. Use this Skill to jump straight to reaper_loop in src/server/server.cpp, check the root-gone invariant, and run ServerReaperTest.RootDeletionStopsServer. ## Quick Start Ask the AI to use the lci-server-lifecycle skill to trace how lci shutdown --all finds and stops running index servers.
