@candor.sh/cli 0.2.0 → 0.3.0

package/dist/index.cjs

@@ -4484,7 +4484,7 @@ var import_os = require("os");
 var CANDOR_DIR = (0, import_path.join)((0, import_os.homedir)(), ".candor");
 var CONFIG_PATH = (0, import_path.join)(CANDOR_DIR, "config.json");
 var DEFAULT_CONFIG = {
-  apiUrl: "https://api.candor.sh"
+  apiUrl: "https://candor.sh"
 };
 function ensureCandorDir() {
   if (!(0, import_fs.existsSync)(CANDOR_DIR)) {
@@ -4513,7 +4513,7 @@ function getConfigPath() {
 }
 
 // src/skill-content.ts
-var SKILL_CONTENT = '# Candor \u2014 AI-Moderated User Studies\n\nCandor lets you create and manage AI-moderated user studies directly from the terminal. An AI voice agent interviews real participants while they use your product, then delivers prioritized findings.\n\n## Available Commands\n\nRun these via the Bash tool. Always use `--json` for machine-readable output.\n\n- `candor study create --url <url> --goal "<goal>" [--participants N] --json` \u2014 Create a study\n- `candor study list --json` \u2014 List all studies (or `candor study --json`)\n- `candor study status <study-id> --json` \u2014 Check participant progress\n- `candor study findings <study-id> --json` \u2014 Get prioritized feedback (P0\u2013P3)\n- `candor study approve <study-id> --json` \u2014 Approve recruitment (human-in-the-loop gate)\n\n## Workflow\n\n### Creating a Study\n\n1. **Infer context** \u2014 If the user doesn\'t provide a URL, check:\n   - `package.json` `homepage` field\n   - README for a deployed URL\n   - Ask the user if nothing is found\n2. **Run `candor study create`** with the product URL and goal\n3. **Show the study summary** including:\n   - Drafted study script (sections, questions, tasks)\n   - Participant count and demographic filters\n   - Estimated cost per session\n4. **Wait for explicit approval** \u2014 Do NOT run `candor study approve` without the user saying "yes", "approve", "go ahead", or similar\n5. **On approval**, run `candor study approve <study-id>` to begin recruiting\n\n### Checking Results\n\n- Use `candor study status <id> --json` to show completion progress\n- Use `candor study findings <id> --json` to show prioritized findings\n- Offer to create GitHub issues for P0/P1 items if the user wants\n\n## Important Rules\n\n- NEVER approve a study without explicit human confirmation\n- ALWAYS show cost estimate before approval\n- Studies are billed per participant session (~$14.50/session)\n- If Candor is not initialized, tell the user to run `candor init` first\n';
+var SKILL_CONTENT = '# Candor \u2014 AI-Moderated User Studies & Human Evaluation\n\nCandor lets you create and manage AI-moderated user studies and human evaluation jobs directly from the terminal. Use `candor study` for product research (AI voice interviews with real users) and `candor eval` for quick human labeling, ranking, and evaluation tasks via Mechanical Turk.\n\n## Available Commands\n\nRun these via the Bash tool. Always use `--json` for machine-readable output.\n\n### Studies (product research)\n\n- `candor study create --url <url> --goal "<goal>" [--participants N] --json` \u2014 Create a study\n- `candor study list --json` \u2014 List all studies (or `candor study --json`)\n- `candor study status <study-id> --json` \u2014 Check participant progress\n- `candor study findings <study-id> --json` \u2014 Get prioritized feedback (P0\u2013P3)\n- `candor study approve <study-id> --json` \u2014 Approve recruitment (human-in-the-loop gate)\n\n### Evals (quick human evaluation / labeling)\n\n- `candor eval create --goal "<goal>" --items "<item1,item2,...>" [--type pairwise_comparison] [--workers N] [--reward <cents>] --json` \u2014 Create an eval\n- `candor eval list --json` \u2014 List all evals\n- `candor eval approve <eval-id> --json` \u2014 Launch eval on MTurk\n- `candor eval status <eval-id> --json` \u2014 Check eval progress\n- `candor eval results <eval-id> --json` \u2014 View results and rankings\n\n## Workflow\n\n### Creating a Study\n\n1. **Infer context** \u2014 If the user doesn\'t provide a URL, check:\n   - `package.json` `homepage` field\n   - README for a deployed URL\n   - Ask the user if nothing is found\n2. **Run `candor study create`** with the product URL and goal\n3. **Show the study summary** including:\n   - Drafted study script (sections, questions, tasks)\n   - Participant count and demographic filters\n   - Estimated cost per session\n4. **Wait for explicit approval** \u2014 Do NOT run `candor study approve` without the user saying "yes", "approve", "go ahead", or similar\n5. **On approval**, run `candor study approve <study-id>` to begin recruiting\n\n### Creating an Eval\n\nEvals are for quick human evaluation tasks like labeling, ranking, rating, or comparing items. They run on Mechanical Turk and return results in minutes.\n\n**When to use eval vs study:**\n- Use `candor eval` when the user wants to label, rank, rate, compare, or categorize items (audio files, images, text, etc.)\n- Use `candor study` when the user wants qualitative product feedback from real users\n\n**Choosing the right task type:**\n- `pairwise_comparison` \u2014 Best for ranking. Workers compare pairs and pick a winner. Produces a stack ranking with win rates. Use when the user says "rank", "compare", "which is better", "stack rank".\n- `categorical_label` \u2014 Workers assign a label from a set of categories. Use when the user says "label", "categorize", "classify", "tag".\n- `rating_scale` \u2014 Workers rate items on a numeric scale. Use when the user says "rate", "score", "quality score".\n- `free_text` \u2014 Workers provide open-ended text feedback.\n\n**Workflow:**\n1. **Identify the items** \u2014 Look for files the user mentions (audio, images, etc.). Extract labels from filenames.\n2. **Infer the task type** \u2014 Based on the user\'s goal, pick the most appropriate type. If vague (e.g. "evaluate these"), default to `pairwise_comparison` for ranking or `rating_scale` for quality assessment.\n3. **Design the eval** \u2014 The API auto-generates the experiment design (randomized pairs, worker counts, etc.) from the goal and items. If the user is prescriptive about methodology, pass their preferences.\n4. **Create the eval** \u2014 Run `candor eval create` with goal and items.\n5. **Show the plan** \u2014 Display task type, item count, pair count, estimated cost and time.\n6. **Wait for approval** \u2014 Do NOT launch without explicit confirmation.\n7. **Launch** \u2014 Run `candor eval approve <id>` to create HITs on MTurk.\n8. **Poll and display results** \u2014 Use `candor eval status <id>` and `candor eval results <id>` to show progress and rankings.\n\n**Example natural language \u2192 eval mapping:**\n\n| User says | Task type | Items |\n|-----------|-----------|-------|\n| "rank these audio samples by TTS quality" | pairwise_comparison | Audio file names |\n| "label these images as cat or dog" | categorical_label | Image file names |\n| "rate these UI mockups on a 1-5 scale" | rating_scale | Mockup file names |\n| "which of these headlines is more engaging" | pairwise_comparison | Headline text |\n| "evaluate these samples" (vague) | pairwise_comparison | File names |\n\n### Checking Results\n\n- Use `candor study status <id> --json` to show completion progress\n- Use `candor study findings <id> --json` to show prioritized findings\n- Use `candor eval status <id> --json` to show eval progress\n- Use `candor eval results <id> --json` to show rankings and agreement metrics\n- Offer to create GitHub issues for P0/P1 items if the user wants\n\n## Important Rules\n\n- NEVER approve a study or launch an eval without explicit human confirmation\n- ALWAYS show cost estimate before launching\n- Studies are billed per participant session (~$14.50/session)\n- Evals are billed per task assignment (~$0.05-0.10/task + 20-40% MTurk fee)\n- If Candor is not initialized, tell the user to run `candor init` first\n';
 
 // src/commands/init.ts
 var CLAUDE_SKILLS_DIR = (0, import_path2.join)((0, import_os2.homedir)(), ".claude", "skills", "candor");
@@ -4970,28 +4970,437 @@ async function updateCommand() {
   }
 }
 function getCurrentVersion() {
+  if ("0.3.0") {
+    return "0.3.0";
+  }
   try {
     const pkgPath = (0, import_path4.join)((0, import_path4.dirname)(new URL(import_meta2.url).pathname), "..", "package.json");
     const pkg = JSON.parse((0, import_fs4.readFileSync)(pkgPath, "utf-8"));
     return pkg.version;
   } catch {
-    return "0.1.0";
+    return "unknown";
+  }
+}
+
+// src/commands/eval-create.ts
+async function evalCreateCommand(options) {
+  const chalk2 = (await Promise.resolve().then(() => (init_source(), source_exports))).default;
+  try {
+    const itemLabels = options.items.split(",").map((s) => s.trim()).filter(Boolean);
+    if (itemLabels.length < 2) {
+      console.log(chalk2.red("Error: At least 2 items are required (comma-separated)"));
+      process.exit(1);
+    }
+    const body = {
+      goal: options.goal,
+      items: itemLabels.map((label) => ({ label }))
+    };
+    if (options.type) body.taskType = options.type;
+    if (options.labels) {
+      const labels = options.labels.split(",").map((s) => s.trim()).filter(Boolean);
+      body.design = { labels };
+    }
+    if (options.workers) body.workersPerTask = parseInt(options.workers, 10);
+    if (options.reward) body.rewardCentsPerTask = parseInt(options.reward, 10);
+    if (options.batchSize) body.batchSize = parseInt(options.batchSize, 10);
+    const data = await apiRequest("/api/evals", { method: "POST", body });
+    if (options.json) {
+      console.log(JSON.stringify(data, null, 2));
+      return;
+    }
+    const result = data;
+    console.log();
+    console.log(chalk2.green(`  + Eval created: ${result.eval.id}`));
+    console.log(chalk2.bold(`  ${result.eval.name}`));
+    console.log(
+      chalk2.dim(
+        `  Type: ${result.eval.taskType} | Workers/task: ${result.eval.workersPerTask}`
+      )
+    );
+    console.log(
+      chalk2.dim(`  Items: ${result.items.length} | Tasks: ${result.estimate.totalTasks}`)
+    );
+    console.log(
+      chalk2.dim(
+        `  Est. cost: $${(result.estimate.estimatedCostCents / 100).toFixed(2)} | Est. time: ~${result.estimate.estimatedMinutes} min`
+      )
+    );
+    console.log(chalk2.dim(`  Status: ${result.eval.status}`));
+    console.log();
+    console.log(
+      chalk2.dim(`  To approve: candor eval approve ${result.eval.id}`)
+    );
+    console.log();
+  } catch (err) {
+    if (options.json) {
+      console.log(
+        JSON.stringify({ error: err instanceof Error ? err.message : String(err) })
+      );
+      process.exit(1);
+    }
+    console.log(chalk2.red(`Error: ${err instanceof Error ? err.message : err}`));
+    process.exit(1);
+  }
+}
+
+// src/commands/eval-launch.ts
+async function evalLaunchCommand(id, options) {
+  const chalk2 = (await Promise.resolve().then(() => (init_source(), source_exports))).default;
+  try {
+    const data = await apiRequest(`/api/evals/${id}/launch`, { method: "POST" });
+    if (options.json) {
+      console.log(JSON.stringify(data, null, 2));
+      return;
+    }
+    const result = data;
+    console.log();
+    console.log(chalk2.green(`  \u26A1 Eval launched!`));
+    console.log(chalk2.dim(`  ${result.totalTasks} tasks being sent to workers`));
+    console.log(chalk2.dim(`  Results will start streaming in within minutes.`));
+    console.log();
+    console.log(chalk2.dim(`  Check progress: candor eval status ${id}`));
+    console.log(chalk2.dim(`  View results:   candor eval results ${id}`));
+    console.log();
+  } catch (err) {
+    if (options.json) {
+      console.log(
+        JSON.stringify({ error: err instanceof Error ? err.message : String(err) })
+      );
+      process.exit(1);
+    }
+    console.log(chalk2.red(`Error: ${err instanceof Error ? err.message : err}`));
+    process.exit(1);
+  }
+}
+
+// src/commands/eval-status.ts
+function formatTime(iso) {
+  const d = new Date(iso);
+  return d.toLocaleTimeString([], { hour: "2-digit", minute: "2-digit", second: "2-digit" });
+}
+function renderStatus(result, id, chalk2, opts = {}) {
+  const p = result.progress;
+  const expectedResponses = p.totalTasks * p.workersPerTask;
+  const responsePct = expectedResponses > 0 ? Math.round(p.totalResponses / expectedResponses * 100) : 0;
+  const barLen = 20;
+  const filled = Math.round(responsePct / 100 * barLen);
+  const bar = "\u2588".repeat(filled) + "\u2591".repeat(barLen - filled);
+  const lines = [];
+  lines.push("");
+  lines.push(chalk2.bold(`  ${result.eval.name}`));
+  lines.push(
+    chalk2.dim(`  ${result.eval.taskType} | ${result.eval.status}`)
+  );
+  lines.push("");
+  lines.push(`  [${bar}] ${responsePct}%`);
+  lines.push(
+    chalk2.dim(
+      `  ${p.totalResponses}/${expectedResponses} responses (${p.uniqueWorkers}/${p.workersPerTask} workers \xD7 ${p.totalTasks} tasks)`
+    )
+  );
+  if (opts.showActivity && result.activity.length > 0) {
+    lines.push("");
+    lines.push(chalk2.bold("  Activity"));
+    const entries = result.activity;
+    const seenCount = opts.seenActivityCount ?? 0;
+    for (let i = 0; i < entries.length; i++) {
+      const a = entries[i];
+      const time = chalk2.dim(formatTime(a.at));
+      const isNew = i >= seenCount;
+      const msg = isNew ? chalk2.green(a.message) : chalk2.white(a.message);
+      lines.push(`  ${time}  ${msg}`);
+    }
+  }
+  lines.push("");
+  if (result.eval.status === "completed") {
+    lines.push(chalk2.green("  \u2713 Eval complete!"));
+    lines.push(chalk2.dim(`  View results: candor eval results ${id}`));
+    lines.push("");
+  } else if (!opts.showActivity && result.eval.status === "active") {
+    lines.push(chalk2.dim(`  Preview results: candor eval results ${id}`));
+    lines.push("");
+  }
+  return lines.join("\n");
+}
+async function evalStatusCommand(id, options) {
+  const chalk2 = (await Promise.resolve().then(() => (init_source(), source_exports))).default;
+  const fetchStatus = async () => {
+    const data = await apiRequest(`/api/evals/${id}`);
+    return data;
+  };
+  try {
+    if (options.json && !options.live) {
+      const data = await fetchStatus();
+      console.log(JSON.stringify(data, null, 2));
+      return;
+    }
+    if (!options.live) {
+      const result = await fetchStatus();
+      console.log(renderStatus(result, id, chalk2, { showActivity: result.activity?.length > 0 }));
+      return;
+    }
+    let lastResponseCount = -1;
+    let seenActivityCount = 0;
+    const poll = async () => {
+      const result = await fetchStatus();
+      if (options.json) {
+        if (result.progress.totalResponses !== lastResponseCount) {
+          console.log(JSON.stringify(result));
+          lastResponseCount = result.progress.totalResponses;
+        }
+      } else {
+        process.stdout.write("\x1B[H\x1B[J");
+        const prevSeen = seenActivityCount;
+        seenActivityCount = result.activity?.length ?? 0;
+        console.log(
+          renderStatus(result, id, chalk2, {
+            showActivity: true,
+            seenActivityCount: prevSeen
+          })
+        );
+        lastResponseCount = result.progress.totalResponses;
+        if (result.eval.status !== "completed") {
+          console.log(chalk2.dim("  Polling every 5s... Ctrl+C to stop"));
+        }
+      }
+      return result.eval.status === "completed";
+    };
+    const done = await poll();
+    if (done) return;
+    await new Promise((resolve) => {
+      const interval = setInterval(async () => {
+        try {
+          const done2 = await poll();
+          if (done2) {
+            clearInterval(interval);
+            resolve();
+          }
+        } catch {
+          clearInterval(interval);
+          resolve();
+        }
+      }, 5e3);
+      process.on("SIGINT", () => {
+        clearInterval(interval);
+        console.log("");
+        resolve();
+      });
+    });
+  } catch (err) {
+    if (options.json) {
+      console.log(
+        JSON.stringify({
+          error: err instanceof Error ? err.message : String(err)
+        })
+      );
+      process.exit(1);
+    }
+    console.log(
+      chalk2.red(`Error: ${err instanceof Error ? err.message : err}`)
+    );
+    process.exit(1);
+  }
+}
+
+// src/commands/eval-results.ts
+async function evalResultsCommand(id, options) {
+  const chalk2 = (await Promise.resolve().then(() => (init_source(), source_exports))).default;
+  try {
+    const data = await apiRequest(`/api/evals/${id}/results`);
+    if (options.json) {
+      console.log(JSON.stringify(data, null, 2));
+      return;
+    }
+    const result = data;
+    if (!result.results) {
+      console.log();
+      console.log(chalk2.yellow(`  No results yet.`));
+      if (result.message) console.log(chalk2.dim(`  ${result.message}`));
+      console.log();
+      return;
+    }
+    console.log();
+    if ("rankings" in result.results) {
+      const pairwise = result.results;
+      console.log(chalk2.bold("  Rankings"));
+      console.log(chalk2.dim("  \u2500".repeat(30)));
+      for (const r of pairwise.rankings) {
+        const winPct = (r.winRate * 100).toFixed(1);
+        const barLen = 15;
+        const filled = Math.round(r.winRate * barLen);
+        const bar = "\u2588".repeat(filled) + "\u2591".repeat(barLen - filled);
+        console.log(
+          `  ${chalk2.bold(`#${r.rank}`)} ${r.label.padEnd(25)} [${bar}] ${winPct}%  (${r.totalWins}/${r.totalComparisons} wins)`
+        );
+      }
+      console.log();
+      console.log(chalk2.bold("  Agreement Metrics"));
+      console.log(chalk2.dim("  \u2500".repeat(30)));
+      const agr = pairwise.agreement;
+      console.log(
+        `  Pairwise agreement: ${(agr.pairwiseAgreementRate * 100).toFixed(1)}%`
+      );
+      console.log(`  Krippendorff's \u03B1:  ${agr.krippendorphAlpha.toFixed(3)}`);
+      if (agr.krippendorphAlpha >= 0.8) {
+        console.log(chalk2.green("  \u2192 Strong agreement"));
+      } else if (agr.krippendorphAlpha >= 0.667) {
+        console.log(chalk2.yellow("  \u2192 Moderate agreement"));
+      } else {
+        console.log(chalk2.red("  \u2192 Low agreement \u2014 consider more workers per task"));
+      }
+      if (agr.disagreedPairs.length > 0) {
+        console.log();
+        console.log(
+          chalk2.dim(
+            `  ${agr.disagreedPairs.length} pairs with disagreement`
+          )
+        );
+      }
+      console.log();
+      console.log(
+        chalk2.dim(
+          `  ${pairwise.totalResponses} responses across ${pairwise.totalPairs} pairs`
+        )
+      );
+    }
+    if ("items" in result.results && !("rankings" in result.results)) {
+      const categorical = result.results;
+      console.log(chalk2.bold("  Labels"));
+      console.log(chalk2.dim("  \u2500".repeat(30)));
+      for (const item of categorical.items) {
+        const conf = (item.confidence * 100).toFixed(0);
+        const distStr = Object.entries(item.labelDistribution).map(([k, v]) => `${k}:${v}`).join(" ");
+        console.log(
+          `  ${item.label.padEnd(25)} \u2192 ${chalk2.bold(item.assignedLabel)} (${conf}% confidence)  [${distStr}]`
+        );
+      }
+    }
+    console.log();
+    if (result.progress) {
+      const p = result.progress;
+      if (result.status === "completed") {
+        console.log(
+          chalk2.dim(
+            `  ${p.totalResponses} responses from ${p.uniqueWorkers || "?"} workers`
+          )
+        );
+      } else {
+        const workers = p.uniqueWorkers != null ? `${p.uniqueWorkers}/${p.workersPerTask || "?"}` : "?";
+        console.log(
+          chalk2.dim(
+            `  ${p.totalResponses} responses (${workers} workers) \u2014 results are partial`
+          )
+        );
+      }
+      console.log();
+    }
+  } catch (err) {
+    if (options.json) {
+      console.log(
+        JSON.stringify({ error: err instanceof Error ? err.message : String(err) })
+      );
+      process.exit(1);
+    }
+    console.log(chalk2.red(`Error: ${err instanceof Error ? err.message : err}`));
+    process.exit(1);
+  }
+}
+
+// src/commands/eval-list.ts
+async function evalListCommand(options) {
+  const chalk2 = (await Promise.resolve().then(() => (init_source(), source_exports))).default;
+  try {
+    const data = await apiRequest("/api/evals");
+    if (options.json) {
+      console.log(JSON.stringify(data, null, 2));
+      return;
+    }
+    const result = data;
+    if (result.evals.length === 0) {
+      console.log();
+      console.log(chalk2.dim("  No evals yet. Create one with:"));
+      console.log(chalk2.dim('  candor eval create --goal "rank by quality" --items "a,b,c"'));
+      console.log();
+      return;
+    }
+    console.log();
+    for (const e of result.evals) {
+      const statusColor = e.status === "completed" ? chalk2.green : e.status === "active" ? chalk2.yellow : chalk2.dim;
+      const progress = e.totalTasks ? ` (${e.completedTasks ?? 0}/${e.totalTasks} done)` : "";
+      console.log(`  ${chalk2.bold(e.id)}  ${e.name}`);
+      console.log(
+        chalk2.dim(`  ${e.taskType} | ${statusColor(e.status)}${progress}`)
+      );
+      console.log();
+    }
+  } catch (err) {
+    if (options.json) {
+      console.log(
+        JSON.stringify({ error: err instanceof Error ? err.message : String(err) })
+      );
+      process.exit(1);
+    }
+    console.log(chalk2.red(`Error: ${err instanceof Error ? err.message : err}`));
+    process.exit(1);
+  }
+}
+
+// src/commands/eval-cancel.ts
+async function evalCancelCommand(id, options) {
+  const chalk2 = (await Promise.resolve().then(() => (init_source(), source_exports))).default;
+  try {
+    const data = await apiRequest(`/api/evals/${id}/cancel`, {
+      method: "POST"
+    });
+    if (options.json) {
+      console.log(JSON.stringify(data, null, 2));
+      return;
+    }
+    console.log();
+    console.log(chalk2.yellow(`  Eval ${id} cancelled`));
+    console.log(chalk2.dim(`  Stopped recruiting \u2014 no new workers will be assigned`));
+    if (data.warnings?.length) {
+      console.log();
+      for (const w of data.warnings) {
+        console.log(chalk2.red(`  Warning: ${w}`));
+      }
+    }
+    console.log();
+    console.log(chalk2.dim(`  Any responses already collected are still available:`));
+    console.log(chalk2.dim(`  candor eval results ${id}`));
+    console.log();
+  } catch (err) {
+    if (options.json) {
+      console.log(
+        JSON.stringify({ error: err instanceof Error ? err.message : String(err) })
+      );
+      process.exit(1);
+    }
+    console.log(chalk2.red(`Error: ${err instanceof Error ? err.message : err}`));
+    process.exit(1);
   }
 }
 
 // src/index.ts
 var program2 = new Command();
-program2.name("candor").description("AI-moderated user studies, wired into your dev workflow").version("0.2.0");
+program2.name("candor").description("AI-moderated user studies, wired into your dev workflow").version("0.3.0").enablePositionalOptions();
 program2.command("init").description("Set up Candor: authenticate and configure Claude Code integration").option("--skill-only", "Only reinstall the skill file").action(initCommand);
 program2.command("login").description("Re-authenticate with Candor").action(loginCommand);
 program2.command("logout").description("Log out and clear stored credentials").action(logoutCommand);
-var study = program2.command("study").description("Create, view, and manage user studies").option("--json", "Output as JSON").action(studiesCommand);
+var study = program2.command("study").description("Create, view, and manage user studies").passThroughOptions().option("--json", "Output as JSON").action(studiesCommand);
 study.command("list").description("List all studies").option("--json", "Output as JSON").action(studiesCommand);
 study.command("show <id>").description("View study details and findings").option("--json", "Output as JSON").action(studyCommand);
 study.command("create").description("Create a new user study").requiredOption("--url <url>", "Product URL to test").requiredOption("--goal <goal>", "What you want to learn").option("--participants <n>", "Number of participants", "5").option("--json", "Output as JSON").action(createStudyCommand);
 study.command("status <id>").description("Check participant progress").option("--json", "Output as JSON").action(statusCommand);
 study.command("findings <id>").description("Get prioritized findings").option("--json", "Output as JSON").action(findingsCommand);
 study.command("approve <id>").description("Approve to begin recruiting").option("--json", "Output as JSON").action(approveCommand);
+var evalCmd = program2.command("eval").description("Create, approve, and view human evaluation jobs").passThroughOptions().option("--json", "Output as JSON").action(evalListCommand);
+evalCmd.command("list").description("List all eval jobs").option("--json", "Output as JSON").action(evalListCommand);
+evalCmd.command("create").description("Create a new eval job").requiredOption("--goal <goal>", "What you want to evaluate").requiredOption("--items <items>", "Comma-separated item labels").option("--type <type>", "Task type: pairwise_comparison, categorical_label, rating_scale").option("--labels <labels>", "Comma-separated labels for categorical_label type").option("--workers <n>", "Workers per task", "3").option("--reward <cents>", "Reward in cents per task", "5").option("--batch-size <n>", "Tasks per assignment (auto if not set)").option("--json", "Output as JSON").action(evalCreateCommand);
+evalCmd.command("approve <id>").description("Approve and launch an eval").option("--json", "Output as JSON").action(evalLaunchCommand);
+evalCmd.command("status <id>").description("Check eval progress").option("--json", "Output as JSON").option("--live", "Poll for updates until complete").action(evalStatusCommand);
+evalCmd.command("results <id>").description("View eval results and rankings").option("--json", "Output as JSON").action(evalResultsCommand);
+evalCmd.command("cancel <id>").description("Cancel an active eval and stop recruiting").option("--json", "Output as JSON").action(evalCancelCommand);
 program2.command("update").description("Update the Candor CLI to the latest version").action(updateCommand);
 program2.command("doctor", { hidden: true }).description("Run diagnostic checks on your Candor installation").action(doctorCommand);
 program2.parse();

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@candor.sh/cli",
-  "version": "0.2.0",
+  "version": "0.3.0",
   "private": false,
   "type": "module",
   "bin": {