feat(mcp)!: rename workbench_info → get_workbench_info (TDQS naming) — v2.0.0

AGENTS.md

@@ -13,7 +13,7 @@ offloaded to a compute box (`cs run "pnpm test"`) or CI — do not run them loca
 ## Repo structure
 
 - `mcp-server/` — the `opencode-workbench` MCP server + connector (TypeScript).
-  Tools: `workbench_info`, `inspect_target`, `plan_clone`, `apply_clone`,
+  Tools: `get_workbench_info`, `inspect_target`, `plan_clone`, `apply_clone`,
   `verify_clone`, `install_component`, `list_required_credentials`,
   `run_auth_flow`, `bootstrap_host`. Local + SSH transports.
   See `mcp-server/README.md` and `docs/architecture/clone-mcp.md`.

connector.json

@@ -3,7 +3,7 @@
   "name": "opencode-workbench",
   "title": "OpenCode Workbench",
   "description": "Clones the OpenCode workbench installation onto Linux machines, locally or over SSH.",
-  "version": "1.1.4",
+  "version": "2.0.0",
   "transport": {
     "type": "stdio",
     "command": "npx",

context7.json

@@ -9,7 +9,7 @@
   ],
   "rules": [
     "Run locally with `npx -y @simonmak-ascent/opencode-workbench` (Node stdio MCP server; no API key).",
-    "Drive a clone with the tool order `workbench_info` → `inspect_target` → `plan_clone` → `apply_clone` → `verify_clone`; use `install_component` to install a single component by id, and `bootstrap_host` for a one-call end-to-end provision (platform scan + OpenCode + apply + verify; `help:true` documents parameters).",
+    "Drive a clone with the tool order `get_workbench_info` → `inspect_target` → `plan_clone` → `apply_clone` → `verify_clone`; use `install_component` to install a single component by id, and `bootstrap_host` for a one-call end-to-end provision (platform scan + OpenCode + apply + verify; `help:true` documents parameters).",
     "`apply_clone` and `install_component` are consent-gated: without `confirm:true` they return a plan and change nothing.",
     "Every tool takes a `target` of `{mode: \"local\"}` or `{mode: \"ssh\", host, user, port?, identityFile?}`.",
     "The connector never reads, transmits or writes secret values; `apply_clone` writes an empty `~/.env.workbench` template (mode 600) listing the `{env:VAR}` names to fill in.",

docs/CHANGELOG_WORKBENCH.md

@@ -3,6 +3,19 @@
 > All notable changes to the workbench configuration.
 > Format: [Keep a Changelog](https://keepachangelog.com/en/1.0.0/)
 
+## [2026-10-03] — v2.0.0
+
+### Breaking
+- Renamed the MCP tool `workbench_info` → `get_workbench_info` for a consistent
+  `verb_noun` naming pattern (TDQS coherence). Clients that call `workbench_info`
+  must update. All other tools are unchanged.
+
+### Changed
+- TDQS description pass across all 9 tools: explicit purpose, usage (use/do-not-use),
+  behavioral transparency, and sibling disambiguation.
+- Official MCP Registry: added `mcpName` to `package.json`, `server.json`
+  description within the 100-char limit, workflow uses `login github-oidc`.
+
 ## [2026-10-03] — VDD-ready, key-resilient profile
 
 ### Added

docs/architecture/clone-mcp.md

@@ -37,7 +37,7 @@ mcp-server/src/
 
 | Tool | Purpose |
 |------|---------|
-| `workbench_info` | Repo, components by tier, optional MCP add-ons. |
+| `get_workbench_info` | Repo, components by tier, optional MCP add-ons. |
 | `inspect_target` | OS/arch/pkg-manager/Node/Docker/OpenCode and the *names* of credentials present. |
 | `plan_clone` | Diff the target against the profile → steps (with privileged-command previews). |
 | `apply_clone` | Consent-gated: without `confirm:true` returns the plan; with it, clones + installs. |

docs/architecture/opencode-runtime-snapshot.md

@@ -35,7 +35,7 @@ Notes:
 
 ## MCP Tool Surface
 
-The server exposes **9 tools**: `workbench_info`, `inspect_target`, `plan_clone`,
+The server exposes **9 tools**: `get_workbench_info`, `inspect_target`, `plan_clone`,
 `apply_clone`, `verify_clone`, `install_component`, `list_required_credentials`,
 `run_auth_flow`, `bootstrap_host`.
 

llms.txt

@@ -15,7 +15,7 @@
 
 ## MCP tools (9)
 
-- `workbench_info` — describe this server: repo, components by tier, optional MCP add-ons
+- `get_workbench_info` — describe this server: repo, components by tier, optional MCP add-ons
 - `inspect_target` — probe a machine: OS, arch, package manager, Node, Docker, OpenCode, home and npm-global paths
 - `plan_clone` — diff the target against the profile → install / present / manual steps (with command previews)
 - `apply_clone` — consent-gated: without `confirm:true` returns the plan; with it, clones + installs the profile (idempotent)

mcp-server/README.md

@@ -34,7 +34,7 @@ From a checkout, use the local build instead:
 
 | Tool | Purpose |
 |------|---------|
-| `workbench_info` | Repo, components by tier, optional MCP add-ons. |
+| `get_workbench_info` | Repo, components by tier, optional MCP add-ons. |
 | `inspect_target` | OS/arch/pkg-manager/Node/Docker/OpenCode/home/npm paths. |
 | `plan_clone` | Diff the target against the profile → steps to install / present / manual. |
 | `apply_clone` | Clone the profile + install missing components + write rendered config. Idempotent. |

mcp-server/src/tools.ts

@@ -199,7 +199,7 @@ function fail(message: string, error: unknown) {
 
 export function registerTools(server: McpServer): void {
   server.registerTool(
-    "workbench_info",
+    "get_workbench_info",
     {
       title: "Workbench server info",
       description:
@@ -323,10 +323,10 @@ export function registerTools(server: McpServer): void {
     {
       title: "Install one Workbench component",
       description:
-        "Install a single Workbench component by id (e.g. `node`, `opencode`, `npm-mcps`, `skills`) on a target; `component` must be an id from workbench_info. Idempotent and consent-gated: without `confirm:true` it returns the plan. Use it for one targeted component; use apply_clone for a component subset, or bootstrap_host for a first-time end-to-end provision — do not combine them for the same host and change.",
+        "Install a single Workbench component by id (e.g. `node`, `opencode`, `npm-mcps`, `skills`) on a target; `component` must be an id from get_workbench_info. Idempotent and consent-gated: without `confirm:true` it returns the plan. Use it for one targeted component; use apply_clone for a component subset, or bootstrap_host for a first-time end-to-end provision — do not combine them for the same host and change.",
       inputSchema: {
         target: targetShape,
-        component: z.string().describe("Component id from workbench_info."),
+        component: z.string().describe("Component id from get_workbench_info."),
         workspace: z.string().optional().describe("Target directory for the profile repo."),
         confirm: z.boolean().optional().describe("Set true to actually install. When absent, the call returns a plan and makes no changes."),
       },

mcp-server/test/server.test.ts

@@ -22,13 +22,13 @@ describe("opencode-workbench server", () => {
     expect(names).toEqual([
       "apply_clone",
       "bootstrap_host",
+      "get_workbench_info",
       "inspect_target",
       "install_component",
       "list_required_credentials",
       "plan_clone",
       "run_auth_flow",
       "verify_clone",
-      "workbench_info",
     ]);
     await client.close();
   });
@@ -47,9 +47,9 @@ describe("opencode-workbench server", () => {
     await client.close();
   });
 
-  it("answers workbench_info with components and optional MCPs", async () => {
+  it("answers get_workbench_info with components and optional MCPs", async () => {
     const client = await connect();
-    const res = await client.callTool({ name: "workbench_info", arguments: {} });
+    const res = await client.callTool({ name: "get_workbench_info", arguments: {} });
     const info = res.structuredContent as {
       components: Array<{ id: string; tier: string }>;
       optionalMcp: string[];

package.json

@@ -1,7 +1,7 @@
 {
   "name": "@simonmak-ascent/opencode-workbench",
   "mcpName": "io.github.simonmak-ascent/opencode-workbench",
-  "version": "1.1.4",
+  "version": "2.0.0",
   "description": "MCP server and connector that clones the OpenCode workbench configuration onto Linux machines (locally or over SSH).",
   "license": "MIT",
   "type": "module",

server.json

@@ -2,7 +2,7 @@
   "$schema": "https://static.modelcontextprotocol.io/schemas/2025-07-09/server.schema.json",
   "name": "io.github.simonmak-ascent/opencode-workbench",
   "description": "Provision an OpenCode workbench and MCP stack on any Linux box, local or over SSH.",
-  "version": "1.1.4",
+  "version": "2.0.0",
   "repository": {
     "url": "https://github.com/simonmak-ascent/opencode-workbench",
     "source": "github"
@@ -11,7 +11,7 @@
     {
       "registryType": "npm",
       "identifier": "@simonmak-ascent/opencode-workbench",
-      "version": "1.1.4",
+      "version": "2.0.0",
       "transport": { "type": "stdio" }
     }
   ],

vdd/specs/mcp-quality-and-registry/spec.md

@@ -53,6 +53,11 @@ the registries my client reads, so I can discover and invoke it correctly.
 - Invent quality scores; only report measured values.
 - Commit registry API keys.
 
+> **Decision (2026-10-03):** the original boundary "never rename tools" was
+> superseded. TDQS flagged `workbench_info` as the lone non-`verb_noun` name, so it
+> was renamed to `get_workbench_info` as a **breaking change in v2.0.0** (recorded
+> in `docs/CHANGELOG_WORKBENCH.md`).
+
 ## Acceptance Criteria
 
 ### AC-1: TDQS lint clean [MUST]

vdd/specs/mcp-quality-and-registry/tdqs-baseline.md

@@ -12,7 +12,7 @@ npx -y mcp-tdqs lint --command 'node mcp-server/dist/index.js' \
 
 | Tool | Params | Coverage | Annotations | Output schema | Cost | Warning |
 |------|--------|----------|-------------|---------------|------|---------|
-| workbench_info | 0 | 100% | yes | yes | 0 | — |
+| workbench_info (now `get_workbench_info`) | 0 | 100% | yes | yes | 0 | — |
 | inspect_target | 1 | 100% | yes | yes | 4 | shadow-candidate (vs bootstrap_host) |
 | plan_clone | 6 | 100% | yes | yes | 4 | shadow-candidate |
 | apply_clone | 7 | 100% | yes | yes | 4 | shadow-candidate |