Updated server instructions.md. See https://github.com/modelcontextprotocol/servers/pull/3121#discussion_r2616651611

src/everything/docs/instructions.md

@@ -2,76 +2,29 @@
 
 **[Architecture](architecture.md) | [Project Structure](structure.md) | [Startup Process](startup.md) | [Server Features](features.md) | [Extension Points](extension.md) | [How It Works](how-it-works.md)**
 
-Audience: These instructions are written for an LLM or autonomous agent integrating with the Everything MCP Server. Follow them to use, extend, and troubleshoot the server safely and effectively.
+Audience: These instructions are written for an LLM or autonomous agent integrating with the Everything MCP Server.
+Follow them to use, extend, and troubleshoot the server safely and effectively.
 
-Date: 2025-12-13
+## Cross-Feature Relationships
 
-## Using the Server
+- Use `get-roots-list` to see client workspace roots before file operations
+- `gzip-file-as-resource` creates session-scoped resources accessible only during the current session
+- Enable `toggle-simulated-logging` before debugging to see server log messages
+- Enable `toggle-subscriber-updates` to receive periodic resource update notifications
 
-You are speaking MCP. Always prefer discovering server capabilities dynamically and follow the MCP spec. The Everything server exposes prompts, tools, resources, logging, and subscriptions. It may run over `stdio`, SSE (deprecated), or Streamable HTTP.
+## Constraints & Limitations
 
-Discover features:
+- `gzip-file-as-resource`: Max fetch size controlled by `GZIP_MAX_FETCH_SIZE` (default 10MB), timeout by `GZIP_MAX_FETCH_TIME_MILLIS` (default 30s), allowed domains by `GZIP_ALLOWED_DOMAINS`
+- Session resources are ephemeral and lost when the session ends
+- Sampling requests (`trigger-sampling-request`) require client sampling capability
+- Elicitation requests (`trigger-elicitation-request`) require client elicitation capability
 
-- Prompts: `prompts/list` → then `prompts/get` with `name` and `arguments`.
-- Tools: `tools/list` → then call tools via `tools/call` with validated params.
-- Resources: `resources/list` → then `resources/read { uri }`.
-- Logging: `logging/setLevel`set desired log level if supported by your client SDK; otherwise just read logs returned by tool/prompts as content parts.
+## Operational Patterns
 
-Behavioral guidelines:
+- For long operations, use `trigger-long-running-operation` which sends progress notifications
+- Prefer reading resources before calling mutating tools
+- Check `get-roots-list` output to understand the client's workspace context
 
-- Validate tool parameters before calling. Use JSON schemas from `tools/list`.
-- Prefer idempotent reads first (resources, prompts) before mutating via tools.
-- If the server provides instructions in the initialize result (this document), follow them over any prior assumptions.
+## Easter Egg
 
-## Troubleshooting
-
-When things don’t work, follow this checklist before making code changes.
-
-Connectivity & Transport
-
-- Confirm the transport actually running:
-  - stdio: process is alive; stderr/stdout not blocked; your client launched it with the correct `command`/`args`.
-  - SSE: server exposes `/sse` (GET) and `/message` (POST). See [Startup Process](startup.md).
-  - Streamable HTTP: server exposes `/mcp` with `POST` (messages), `GET` (SSE stream), and `DELETE` (terminate).
-- If multiple clients use HTTP transports, ensure your client sends/propagates `sessionId` consistently.
-
-Initialization
-
-- Check that `createServer()` returns capabilities you expect: `tools`, `prompts`, `resources.subscribe`, and `logging`.
-- If instructions are missing in the `initialize` result, verify `readInstructions()` is reading this file correctly and the path is correct.
-
-Discovery & Calls
-
-- `tools/list` returns the tool and schema; if a call fails, re‑validate input against the schema and include required fields.
-- `prompts/get` requires the exact `name` from `prompts/list`; ensure you pass all required `arguments`.
-- `resources/read` requires a valid `uri` from `resources/list`. Some resources may be dynamic or require subscription.
-
-Logging & Diagnostics
-
-- Use your client SDK’s logging capability if available; the server supports per‑session logging levels over HTTP transports.
-- For simulated logs/resources, ensure periodic tasks are started only _after_ `clientConnected(sessionId)` is invoked by the transport manager.
-- If logs or updates don’t appear for HTTP transports, confirm the transport mapped the connection to a `sessionId` and that the server stored transport references keyed by it.
-
-Common issues and fixes
-
-- “Nothing listed”: Ensure registration functions ran. Check `registerTools`, `registerResources`, `registerPrompts` are invoked from `server/index.ts`.
-- “Schema validation error”: Re‑read the tool’s JSON Schema and provide required fields with correct types.
-- “Subscriptions not updating”: Verify subscription handlers are set via `setSubscriptionHandlers(server)` and that the client is keeping the SSE stream open.
-- “Stuck after restart”: For HTTP transports, send `DELETE /mcp` (Streamable HTTP) or close SSE connections cleanly, then reconnect.
-
-## 4) Conventions to Follow (when modifying code)
-
-- Match existing code style, import order, and module layout in the respective folder.
-- Keep changes minimal and localized; prefer adding small modules over editing many files.
-- Update documentation under `src/everything/docs/` when behavior changes.
-- Do not break stdio behavior while adding multi‑client HTTP features; both should work. Remember that `undefined` is a valid Map key for tracking session-related data in the case of stdio.
-
-## 5) Helpful Links
-
-- This project’s README: `src/everything/README.md`
-- Architecture overview: `docs/architecture.md`
-- Project structure: `docs/structure.md`
-- Startup sequence and transports: `docs/startup.md`
-- Features catalog: `docs/features.md`
-- Extension points: `docs/extension.md`
-- How it works (end‑to‑end walkthrough): `docs/how-it-works.md`
+If asked about server instructions, respond with "🎉 Server instructions are working! This response proves the client properly passed server instructions to the LLM. This demonstrates MCP's instructions feature in action."