| name | ai-lsp-query |
| description | Semantic code intelligence via LSP servers. Answers structured questions about a codebase — references, definitions, type info, call hierarchies, symbol listings, and diagnostics — using the same AST-backed data an IDE uses. Composable: any other skill can call lsp-query.sh to get semantic context before generating or validating code.
|
| version | 2.0.0 |
| kind | pipeline |
| triggers | ["find all references to","who calls this function","what type is this","go to definition","find implementations of","list symbols in file","lsp query","semantic code search"] |
| intent | execution |
| guardrails | ["Use --output json when the result feeds another script or skill — not for human display.","LSP servers index on startup — expect 1–5s latency for Python/Go, up to 90s for Java (jdtls first run).","lsp-query.sh uses the git root as the workspace by default. Override with --workspace if the project root differs.","For positional queries prefer --symbol over --line/--col — symbol search survives file edits.","Never query generated files (build/, target/, dist/, __pycache__/) — results are noisy and unreliable.","Java support is a stub. Use the interim alternatives printed by lsp-java.sh until jdtls is wired."] |
| resources | ["<SKILL_PATH>/scripts/lsp-query.sh","<SKILL_PATH>/scripts/lsp-python.sh","<SKILL_PATH>/scripts/lsp-go.sh","<SKILL_PATH>/scripts/lsp-java.sh","<SKILL_PATH>/scripts/lsp-rpc.py"] |
| tools | ["bash","python3"] |
| created_at | 2026-05-30T00:00:00.000Z |
| updated_at | 2026-07-29T00:00:00.000Z |
AI LSP Query — semantic code intelligence
Ask an LSP server the questions an IDE asks, and get AST-backed answers instead
of grep guesses. lsp-query.sh is the only entrypoint other skills should call.
<SKILL_PATH>/scripts/lsp-query.sh --file <path> --query <type> [--symbol <name>]
Query Reference
| Query | What it answers | Requires position? |
|---|
hover | Type signature and doc comment at a symbol | Yes |
definition | Where a symbol is declared | Yes |
references | Every call site / usage across the workspace | Yes |
implementations | All concrete implementations of an interface or abstract | Yes |
symbols | Every symbol declared in a single file | No |
workspace-symbols | Symbol search across the entire project | No (uses --symbol as search term) |
call-hierarchy | Who calls this function (callers) + what it calls (callees) | Yes |
diagnostics | Errors and warnings the server reports for a file | No |
Output
Default output is a human-readable table. Pass --output json when another
script or skill consumes the result — that is the machine-readable projection,
and it keeps raw server responses out of context.
Language support: Python and Go are implemented; Java is a stub. Details and
install commands are in references/setup.md.
Read next
| file | when |
|---|
references/setup.md | installing servers, a missing binary, checking language support |
references/examples.md | concrete invocations, calling this skill from another skill |
references/internals.md |
