What problem does it solve? When working on the lci semantic code search engine, it is hard to locate the exact code responsible for how files are scanned, excluded, parsed, and merged into the in-memory index, or to diagnose reindex staleness, watch-mode gaps, and memory growth. This Skill maps the entire indexing pipeline with precise file and line references so you can navigate, debug, and compare it against other code indexers without reading the whole codebase. ## Core Features & Use Cases - Pipeline code map: Entry points, scanner, parser workers, integrator, trigram blooms, content store, and watch pipeline documented with exact source locations. - Config and invariants reference: Every .lci.kdl indexing knob, lock-ordering rules, bulk-window semantics, and known traps such as inert config keys and dead code paths. - Probe recipes and comparisons: Ready commands for auditing excludes, measuring cold index time and RSS, triggering reindex over the unix socket, plus an honest comparison against Zoekt, ripgrep, ctags, and clangd. - Use Case: A user reports that edits are not reflected in search results. Use this Skill to find the watch pipeline code, discover that watch mode silently disables when a .lci.kdl exists without watch_mode true, and verify with the listed tests. ## Quick Start Use the lci-indexing-pipeline skill to find where lci decides which files to exclude from the index and show me how to audit it with the CLI.
