Agent-Skills.md

Agent Skills: Using LSP Tools

Use when mcp-language-server tools are available and you need semantic code intelligence for navigation, refactoring, or type analysis

language-server-protocolsemantic-analysisrefactoringcode-navigationtype-analysis
code-intelligenceID: axiomantic/spellbook/using-lsp-tools

Repository

axiomantic/spellbook
axiomanticLicense: MIT

Install this agent skill to your local

pnpm dlx add-skill https://github.com/axiomantic/spellbook/tree/HEAD/skills/using-lsp-tools

Skill Files

Browse the full folder contents for using-lsp-tools.

Download Skill

Loading file tree…

skills/using-lsp-tools/SKILL.md

Skill Metadata

Name
using-lsp-tools
Description
"Use when mcp-language-server tools are available and you need semantic code intelligence. Triggers: 'find definition', 'find references', 'who calls this', 'rename symbol', 'type hierarchy', 'go to definition', 'where is this used', 'where is this defined', 'what type is this'. Provides navigation, refactoring, and type analysis via LSP."

Using LSP Tools

<ROLE> Language Tooling Expert. Reputation depends on leveraging semantic analysis over text matching for accurate, complete code navigation and refactoring. </ROLE>

Invariant Principles

  1. Semantic > Lexical: LSP understands scope, types, inheritance. Grep sees text.
  2. LSP for Symbols, Grep for Strings: Symbols = definitions, references, types. Strings = TODOs, comments, literals.
  3. Verify Before Fallback: Empty LSP result? Check file saved. Then try text-based.
  4. Atomic Operations Preferred: rename_symbol handles all files. Manual Edit misses references.

Reasoning Schema

<analysis> - Is target a symbol (function, class, variable) or literal text? - Is LSP server active for this language? - Does task need semantic understanding (types, scope, inheritance)? </analysis> <reflection> - Did LSP return expected results? If empty: file saved? Feature supported? - Did fallback find matches LSP missed? Indicates LSP limitation vs. saved state. </reflection>

Inputs

| Input | Required | Description | |-------|----------|-------------| | filePath | Yes | Absolute path to file being analyzed | | line | Context | 1-indexed line number for position-based queries | | column | Context | 1-indexed column for position-based queries | | symbolName | Context | Fully-qualified name for definition/references | | language | No | Language identifier if ambiguous |

Outputs

| Output | Type | Description | |--------|------|-------------| | Symbol locations | Inline | File paths and positions from navigation queries | | Type information | Inline | Hover/signature data for understanding | | Refactoring edits | Applied | Direct code modifications from rename/actions | | Diagnostics | Inline | Errors and warnings for debugging |

Tool Priority Matrix

| Task | LSP Tool | Fallback | |------|----------|----------| | Find definition | definition | Grep func X\|class X\|def X | | Find usages | references | Grep symbol name | | Understand symbol | hover | Read + infer | | Rename | rename_symbol | Multi-file Edit (risky) | | File outline | document_symbols | Grep definitions | | Callers | call_hierarchy incoming | Grep + analyze | | Callees | call_hierarchy outgoing | Read function | | Type hierarchy | type_hierarchy | Grep extends/implements | | Workspace search | workspace_symbol_resolve | Glob + Grep | | Refactorings | code_actions | Manual | | Signature | signature_help | Hover or read | | Diagnostics | diagnostics | Build command | | Format | format_document | Formatter CLI | | Edit by line | edit_file | Built-in Edit |

Parameters

Required: filePath (absolute), line/column (1-indexed), symbolName (fully-qualified for definition/references).

Decision Rules

Use LSP when:

  • Finding true definition (not text match)
  • Refactoring (rename, extract, inline)
  • Understanding type relationships
  • Finding semantic usages
  • Cross-file navigation via imports

Use Grep/Glob when:

  • Literal strings, comments, non-code text
  • Regex patterns
  • LSP returns empty but code exists
  • Unsupported languages
  • Non-symbols (TODOs, URLs, magic strings)

Workflows

Exploration: document_symbols (structure) -> hover (types) -> definition (jump) -> references (usage)

Refactoring: code_actions (discover) -> rename_symbol (execute) OR references (assess impact) -> manual

Type debugging: hover (inferred) -> type_hierarchy (inheritance) -> diagnostics (errors)

Call analysis: call_hierarchy incoming = "who calls?" | outgoing = "what calls?"

Anti-Patterns

<FORBIDDEN> - Using Grep for symbol rename (misses scoped references, hits false positives) - Skipping LSP for "simple" refactors (simple becomes complex with inheritance) - Trusting empty LSP results without checking file saved state - Manual multi-file edits when `rename_symbol` available - Ignoring `diagnostics` output when debugging type errors </FORBIDDEN>

Fallback Protocol

  1. LSP error/empty -> Check file saved (LSP reads disk)
  2. Try table fallback
  3. Persistent failure -> Feature unsupported by server

Self-Check

Before completing:

  • [ ] Used semantic LSP tool for symbol-based queries (not text search)
  • [ ] Verified file saved if LSP returned empty/unexpected results
  • [ ] Applied atomic refactoring operations where available
  • [ ] Documented fallback rationale if LSP bypassed

If ANY unchecked: STOP and reconsider approach.