@przeprogramowani/10x-cli 1.16.0 → 1.18.0

package/README.md

@@ -62,6 +62,7 @@ Once installed, just tell your agent to **set up 10x-cli** and it will pick up t
 | `10x sync` | Bulk-download / refresh lessons and report what changed upstream |
 | `10x doctor` | Diagnose auth, API connectivity, and local config |
 | `10x bench` | Live top-10 AI model leaderboard from [10xbench.ai](https://10xbench.ai) — no login needed |
+| `10x bench-kit <action>` | Create (`init`) and update (`update`) a company benchmark instance from the [10x-bench-kit](https://github.com/przeprogramowani/10x-bench-kit) template |
 
 ### `10x get` Flags
 
@@ -180,6 +181,40 @@ color-coded score bars in your terminal. Public data, works without logging in.
 The data refreshes whenever new benchmark results are published on 10xbench.ai.
 Colors honor `NO_COLOR` and are disabled automatically when output is piped.
 
+### `10x bench-kit`
+
+Creates and maintains a **company benchmark instance** from the
+[10x-bench-kit](https://github.com/przeprogramowani/10x-bench-kit) template —
+a self-hosted benchmark that scores AI agents on tasks embedded in your own
+repositories. The CLI is a thin installer/updater; tasks, assertions, and
+scoring live in the template and the instance.
+
+| Action | Description |
+|--------|-------------|
+| `init [dir]` | Materialize a fresh instance from the template (no git history, fresh `git init`), register the detected base repo, install runner dependencies |
+| `update [dir]` | Upgrade the instance to a newer template: runtime zone replaced wholesale, skills proposed as a reviewable diff, company content untouched |
+
+| Flag | Description |
+|------|-------------|
+| `--template-version <tag>` | Template tag to install (default: latest) |
+| `--tool <id>` | Agent tool for skill placement (`claude-code`, `cursor`, `copilot`, `codex`, `windsurf`, `gemini`, `generic`) |
+| `--yes` | Run non-interactively, accepting defaults |
+
+```bash
+# Create an instance next to your product repo (run inside it to auto-register)
+10x bench-kit init my-benchmark
+
+# Pin the template version
+10x bench-kit init my-benchmark --template-version v0.8.0
+
+# Upgrade an existing instance to the latest template
+10x bench-kit update
+```
+
+Re-running `init` on an existing instance is a **repair**: missing template
+files are restored, company content is never touched. After `update`, run the
+instance's `bench validate` to confirm the content still matches the new kit.
+
 ### Global Flags
 
 - `--json` — Machine-readable JSON output (auto-detected when piped)

package/dist/index.mjs

@@ -10173,7 +10173,7 @@ var cac = (name = "") => new CAC(name);
 // package.json
 var package_default = {
   name: "@przeprogramowani/10x-cli",
-  version: "1.16.0",
+  version: "1.18.0",
   description: "Open-source CLI for 10xDevs course content",
   repository: {
     type: "git",
@@ -11104,19 +11104,6 @@ import {
 import { tmpdir } from "node:os";
 import { basename, join as join4, resolve, sep } from "node:path";
 
-// src/lib/experimental.ts
-var EXPERIMENTAL_ENV = "TENX_CLI_EXPERIMENTAL";
-function experimentalEnabled(env = process.env) {
-  const value = env[EXPERIMENTAL_ENV];
-  return value === "1" || value === "true";
-}
-function requireExperimental(command, flags, env = process.env) {
-  if (experimentalEnabled(env))
-    return;
-  const ctx = resolveContext(flags);
-  outputError(ctx, "experimental_locked", `'10x ${command}' is experimental and currently locked.`, ExitCodes.FORBIDDEN, `Set ${EXPERIMENTAL_ENV}=1 and run '10x ${command}' again.`);
-}
-
 // src/lib/tool-detect.ts
 import { existsSync as existsSync3 } from "node:fs";
 import { join as join3 } from "node:path";
@@ -11462,10 +11449,7 @@ var TEMPLATE_REPO_URL = "https://github.com/przeprogramowani/10x-bench-kit";
 var SHARED_ROOT_FILES = ["AGENTS.md"];
 var PLACEHOLDER_BASE_REPO = "demo-app";
 function registerBenchKitCommand(cli) {
-  if (!experimentalEnabled())
-    return;
-  cli.command("bench-kit <action> [dir]", "Manage a benchmark instance (actions: init, update; experimental)").option("--template-version <tag>", "Template tag to install (default: latest)").option("--tool <id>", `Agent tool for skill placement (${Object.keys(PROFILES).join(", ")})`).option("--yes", "Run non-interactively, accepting defaults").example("10x bench-kit init my-benchmark").example("10x bench-kit init my-benchmark --template-version v0.1.0").example("10x bench-kit update").action(async (action, dir, options) => {
-    requireExperimental(`bench-kit ${action}`, options);
+  cli.command("bench-kit <action> [dir]", "Manage a benchmark instance (actions: init, update)").option("--template-version <tag>", "Template tag to install (default: latest)").option("--tool <id>", `Agent tool for skill placement (${Object.keys(PROFILES).join(", ")})`).option("--yes", "Run non-interactively, accepting defaults").example("10x bench-kit init my-benchmark").example("10x bench-kit init my-benchmark --template-version v0.1.0").example("10x bench-kit update").action(async (action, dir, options) => {
     const ctx = resolveContext(options);
     if (action === "init") {
       await runBenchKitInit(ctx, dir, options);
@@ -11512,15 +11496,27 @@ async function runBenchKitInit(ctx, dirArg, options, deps = defaultDeps) {
     const skills = syncDir(join4(scratch, skillSource), join4(targetDir, skillRootFor(toolId)), { overwrite: !repair });
     installWorkflows(targetDir, { skipExisting: repair });
     let baseRepo = null;
+    let demoTasksPinned = 0;
     if (!repair) {
       const detected = await deps.detectBaseRepo(process.cwd());
       if (detected !== null && resolve(detected.rootDir) !== targetDir) {
-        if (await registerBaseRepo(join4(targetDir, "bench.config.yaml"), detected)) {
-          baseRepo = detected;
-          verbose(ctx, `registered base repo ${detected.name} (${detected.url})`);
+        let repo = detected;
+        const https = toHttpsUrl(detected.url);
+        if (https !== null && await deps.remoteReachable(https)) {
+          repo = { ...detected, url: https };
+          verbose(ctx, `repo answers over https — using ${https} instead of SSH`);
+        }
+        if (await registerBaseRepo(join4(targetDir, "bench.config.yaml"), repo)) {
+          baseRepo = repo;
+          verbose(ctx, `registered base repo ${repo.name} (${repo.url})`);
+          demoTasksPinned = await pinPlaceholderTasks(join4(targetDir, "tasks"), repo);
+          if (demoTasksPinned > 0) {
+            verbose(ctx, `pinned ${demoTasksPinned} demo task(s) to ${repo.headCommit.slice(0, 12)}`);
+          }
         }
       }
     }
+    const runnerDeps = await installRunnerDependencies(ctx, deps, targetDir);
     const manifest = {
       templateVersion,
       templateRef: requestedRef ?? "latest",
@@ -11543,6 +11539,8 @@ async function runBenchKitInit(ctx, dirArg, options, deps = defaultDeps) {
       `Created a benchmark instance in '${targetDir}' from template ${templateVersion}.`,
       `Agent skills installed for ${toolName} under ${skillRootFor(toolId)}/.`,
       baseRepo === null ? "No product repo detected here — add your base repos to bench.config.yaml." : `Registered '${baseRepo.name}' (${baseRepo.url}) as the first base repo in bench.config.yaml.`,
+      ...demoTasksPinned > 0 ? [`Pinned the demo task to ${baseRepo?.headCommit.slice(0, 12)} (current HEAD of the base repo).`] : [],
+      ...runnerDeps === "failed" ? ["Runner dependencies did not install — run 'npm ci --prefix .bench-kit/runner' yourself."] : runnerDeps === "installed" ? ["Runner dependencies installed (.bench-kit/runner/node_modules)."] : [],
       committed ? "Initialized a fresh git repository with an initial commit." : "Initialized a fresh git repository (initial commit skipped — commit the files yourself).",
       "Next: wire up secrets, then run 'bench validate' before the first run."
     ];
@@ -11556,6 +11554,8 @@ async function runBenchKitInit(ctx, dirArg, options, deps = defaultDeps) {
       skillRoot: skillRootFor(toolId),
       filesCopied: copied + skills.added,
       baseRepo,
+      demoTasksPinned,
+      runnerDeps,
       gitInitialized: !repair,
       committed
     });
@@ -11623,9 +11623,11 @@ async function runBenchKitUpdate(ctx, dirArg, options, deps = defaultDeps) {
     for (const file of SHARED_ROOT_FILES) {
       addSync(shared, syncFile(join4(scratch, file), join4(targetDir, file)));
     }
+    const runnerDeps = await installRunnerDependencies(ctx, deps, targetDir);
     output(ctx, [
       `Updated the benchmark instance from template ${currentVersion} to ${newVersion}.`,
       "  .bench-kit/            replaced wholesale (runtime zone)",
+      ...runnerDeps === "failed" ? ["  .bench-kit/runner/     npm ci FAILED — run 'npm ci --prefix .bench-kit/runner' yourself"] : runnerDeps === "installed" ? ["  .bench-kit/runner/     dependencies reinstalled (npm ci)"] : [],
       `  .github/workflows/     ${describeSync(workflows)}`,
       `  ${`${skillRootFor(toolId)}/`.padEnd(23)}${describeSync(skills)} — proposal, review before committing`,
       `  ${SHARED_ROOT_FILES.join(", ").padEnd(23)}${describeSync(shared)} — proposal, review before committing`,
@@ -11641,6 +11643,7 @@ async function runBenchKitUpdate(ctx, dirArg, options, deps = defaultDeps) {
       templateRef: updatedManifest.templateRef,
       tool: toolId,
       skillRoot: skillRootFor(toolId),
+      runnerDeps,
       zones: { benchKit: "replaced", workflows, skills, shared }
     });
   } finally {
@@ -11831,6 +11834,53 @@ async function registerBaseRepo(configPath, repo) {
   writeFileSync3(configPath, doc.toString());
   return true;
 }
+function toHttpsUrl(url) {
+  const scp = url.match(/^git@([^:/]+):(.+)$/);
+  if (scp !== null)
+    return `https://${scp[1]}/${scp[2]}`;
+  const ssh = url.match(/^ssh:\/\/(?:[^@/]+@)?([^:/]+)(?::\d+)?\/(.+)$/);
+  if (ssh !== null)
+    return `https://${ssh[1]}/${ssh[2]}`;
+  return null;
+}
+var PLACEHOLDER_COMMIT = /^0{40}$/;
+async function pinPlaceholderTasks(tasksDir, repo) {
+  if (!existsSync4(tasksDir) || !/^[0-9a-f]{40}$/.test(repo.headCommit))
+    return 0;
+  const { parseDocument } = await Promise.resolve().then(() => __toESM(require_dist(), 1));
+  let pinned = 0;
+  for (const entry of readdirSync(tasksDir, { withFileTypes: true })) {
+    if (!entry.isDirectory())
+      continue;
+    const taskYaml = join4(tasksDir, entry.name, "task.yaml");
+    if (!existsSync4(taskYaml))
+      continue;
+    const doc = parseDocument(readFileSync3(taskYaml, "utf8"));
+    if (doc.getIn(["repo"]) !== PLACEHOLDER_BASE_REPO)
+      continue;
+    const commit = doc.getIn(["commit"]);
+    doc.setIn(["repo"], repo.name);
+    if (typeof commit === "string" && PLACEHOLDER_COMMIT.test(commit)) {
+      doc.setIn(["commit"], repo.headCommit);
+    }
+    writeFileSync3(taskYaml, doc.toString());
+    pinned++;
+  }
+  return pinned;
+}
+async function installRunnerDependencies(ctx, deps, targetDir) {
+  const runnerDir = join4(targetDir, ".bench-kit", "runner");
+  if (!existsSync4(join4(runnerDir, "package.json")))
+    return "skipped";
+  verbose(ctx, "installing runner dependencies (npm ci in .bench-kit/runner)");
+  const result = await deps.installRunnerDeps(runnerDir);
+  if (!result.ok) {
+    verbose(ctx, `npm ci failed (${result.error.trim().split(`
+`).pop() ?? ""})`);
+    return "failed";
+  }
+  return "installed";
+}
 async function freshGitInit(ctx, deps, dir, templateVersion) {
   const init = await deps.runGit(["init"], dir);
   if (!init.ok) {
@@ -11844,9 +11894,13 @@ async function freshGitInit(ctx, deps, dir, templateVersion) {
   }
   return true;
 }
-function run(cmd, args, cwd) {
+function run(cmd, args, cwd, env) {
   return new Promise((resolvePromise) => {
-    const child = spawn(cmd, args, { cwd, stdio: ["ignore", "pipe", "pipe"] });
+    const child = spawn(cmd, args, {
+      cwd,
+      stdio: ["ignore", "pipe", "pipe"],
+      ...env === undefined ? {} : { env: { ...process.env, ...env } }
+    });
     let stdout = "";
     let stderr = "";
     child.stdout?.on("data", (chunk) => {
@@ -11893,6 +11947,15 @@ var defaultDeps = {
       return null;
     return { rootDir, name: basename(rootDir), url, headCommit };
   },
+  async remoteReachable(url) {
+    const result = await run("git", ["ls-remote", "--heads", url], undefined, {
+      GIT_TERMINAL_PROMPT: "0"
+    });
+    return result.ok;
+  },
+  installRunnerDeps(runnerDir) {
+    return run("npm", ["ci", "--no-audit", "--no-fund"], runnerDir);
+  },
   detectToolSignals(cwd) {
     return detectTools(cwd);
   },

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@przeprogramowani/10x-cli",
-  "version": "1.16.0",
+  "version": "1.18.0",
   "description": "Open-source CLI for 10xDevs course content",
   "repository": {
     "type": "git",