core-code-intelligence-lsp

# Code Intelligence (LSP) Grep tells you what string appears. LSP tells you what it **means**. A grep for `UserID` finds every occurrence of those seven characters — including comments, string literals, variable names that happen to share a prefix, and the actual type declaration. An LSP `go-to-definition` query on `UserID` resolves through imports, re-exports, and aliases to the canonical declaration, with its type signature, in milliseconds. That difference is not cosmetic. A refactor grounded in grep assumptions silently breaks code paths that the grep didn't surface. An analysis grounded in LSP diagnostics catches the breakage before it ships. This skill is the cognitive operating manual for using LSP output effectively. It is not a server-installation or DevOps guide — it assumes the LSP server is running and reachable (either natively in the development environment or via an MCP adapter that exposes LSP queries as tools). The doctrine here is: which server to call, which query to issue, how to interpret the result, and what counts as sufficient evidence before making a structural change. ## What This Skill Is (and isn't) **Is:** Doctrine for using LSP output as the primary evidence base during code analysis. Covers query selection, result interpretation, server-health awareness, and the decision rule for when an LSP result is authoritative vs when to read source directly. **Is not:** A guide to installing, configuring, or managing LSP servers. It is not a substitu...