SME Agent Consultation

Before implementing features, fixing bugs, or making architecture changes, you MUST consult the relevant Subject Matter Expert agents in .claude/agents/.

Work Type to Agent Mapping

Work Type	Required Agents
New LSP features	`lsp.md`, `rust-analyzer.md`, `rust.md`
GraphQL validation/linting	`graphql.md`, `apollo-rs.md`
VSCode extension changes	`vscode-extension.md`
CLI tool changes	`graphql-cli.md`
Salsa/incremental computation	`salsa.md`, `rust-analyzer.md`
Salsa debugging (hangs, cache issues)	`salsa.md`
IDE UX features	`graphiql.md`, `lsp.md`
Apollo-specific patterns	`apollo-client.md`, `apollo-rs.md`
Rust API design	`rust.md`

How to Consult

•Identify the work type from the table above
•Read the relevant agent files in .claude/agents/
•Apply the guidance to your implementation
•Document what you learned (see below)

Use the Task tool with subagent_type=general-purpose for deep consultation when needed.

Documentation Requirements

In User Communications

When proposing or explaining a solution, mention which agents were consulted:

"I consulted the lsp.md and rust-analyzer.md agents for guidance on this feature. The LSP agent confirmed this follows the specification, and the rust-analyzer agent recommended using a Salsa query for incremental computation."

In PR Descriptions

Include a "Consulted SME Agents" section:

markdown

## Consulted SME Agents

- **lsp.md**: Confirmed `textDocument/definition` response format
- **rust-analyzer.md**: Recommended query-based architecture for goto definition
- **rust.md**: Advised on error handling patterns using `Result<Option<T>>`

In Issue Comments

Note agent consultations when providing analysis:

"After consulting the graphql.md agent, I can confirm this is expected behavior per section 5.8.3 of the GraphQL specification regarding fragment spread validation."

Why This Matters

•Traceability: Users can understand reasoning behind decisions
•Review Quality: Reviewers know which domain expertise was applied
•Knowledge Transfer: Future sessions can see what guidance was relevant
•Accountability: Ensures agents are actually being consulted

Available Agents

Agent	Domain
`graphql.md`	GraphQL spec compliance, validation rules, type system
`apollo-client.md`	Apollo Client patterns, caching, fragment colocation
`rust-analyzer.md`	Query-based architecture, Salsa, incremental computation
`salsa.md`	Salsa framework, database design, snapshot isolation, concurrency
`rust.md`	Idiomatic Rust, ownership, error handling, API design
`lsp.md`	LSP specification, protocol messages, client compatibility
`graphiql.md`	IDE features, graphql-language-service, UX patterns
`graphql-cli.md`	CLI design, graphql-config, ecosystem tooling
`vscode-extension.md`	Extension development, activation, language client
`apollo-rs.md`	apollo-parser, apollo-compiler, error-tolerant parsing