claude-overnight 1.60.1 → 1.60.2

package/dist/bin/evolve.js

@@ -17,6 +17,8 @@
  */
 import { evolvePrompt } from "../prompt-evolution/index.js";
 import { PLAN_CASES } from "../prompt-evolution/fixtures/plan-cases.js";
+import { STEER_CASES } from "../prompt-evolution/fixtures/steer-cases.js";
+import { COACH_CASES } from "../prompt-evolution/fixtures/coach-cases.js";
 import { harvestRealCases } from "../prompt-evolution/fixtures/harvest.js";
 import { generateCases } from "../prompt-evolution/fixtures/generate.js";
 import { runDiff, runDownload, runPromote } from "./evolve-subcommands.js";
@@ -45,8 +47,8 @@ Options:
   --judge                 Use llm-judge for content scoring (costs extra API calls)
   --judge-model <model>   Model to use for the judge (default: same as eval-model)
   --judge-top-n <n>       Judge only the top-N variants per generation (default: 4)
-  --cases <suite>         Benchmark suite: plan | mcp-planning | mcp-review |
-                          mcp-supervision | mcp-stuck (default: plan)
+  --cases <suite>         Benchmark suite: plan | steer | coach | mcp-planning |
+                          mcp-review | mcp-supervision | mcp-stuck (default: plan)
   --harvest               Append cases harvested from <cwd>/.claude-overnight/runs/*
   --harvest-only          Use ONLY harvested real objectives (fails if none found)
   --harvest-limit <n>     Max harvested cases (default: 10)
@@ -278,6 +280,10 @@ async function evolveOne(opts) {
     else {
         if (opts.cases === "plan")
             cases = opts.harvestOnly ? [] : [...PLAN_CASES];
+        else if (opts.cases === "steer")
+            cases = opts.harvestOnly ? [] : [...STEER_CASES];
+        else if (opts.cases === "coach")
+            cases = opts.harvestOnly ? [] : [...COACH_CASES];
         else
             throw new Error(`Unknown case suite: ${opts.cases}`);
         if (opts.harvest) {

package/dist/core/_version.d.ts

@@ -1 +1 @@
-export declare const VERSION = "1.60.0";
+export declare const VERSION = "1.60.1";

package/dist/core/_version.js

@@ -1,2 +1,2 @@
 // Auto-generated by build — do not edit manually.
-export const VERSION = "1.60.0";
+export const VERSION = "1.60.1";

package/dist/prompt-evolution/fixtures/coach-cases.d.ts

@@ -0,0 +1,2 @@
+import type { BenchmarkCase } from "../types.js";
+export declare const COACH_CASES: BenchmarkCase[];

package/dist/prompt-evolution/fixtures/coach-cases.js

@@ -0,0 +1,64 @@
+export const COACH_CASES = [
+    {
+        name: "simple-todo",
+        hash: "",
+        promptPath: "00_setup/00-1_coach",
+        vars: {
+            objective: "Build a simple todo app using vanilla JS.",
+            tree: "index.html\nstyle.css\napp.js",
+            readme: "# Todo app",
+            providers: "anthropic, openai",
+            isInitialCoach: true,
+        },
+        criteria: {
+            independentTasks: false,
+            specificTasks: false,
+            requiredJsonFields: ["objective", "scope", "recommended", "checklist", "remediation"],
+        },
+    },
+    {
+        name: "vague-objective",
+        hash: "",
+        promptPath: "00_setup/00-1_coach",
+        vars: {
+            objective: "Make it better.",
+            tree: "src/main.ts",
+            readme: "Project",
+            providers: "anthropic",
+            isInitialCoach: true,
+        },
+        criteria: {
+            independentTasks: false,
+            specificTasks: false,
+            requiredJsonFields: ["objective", "scope", "recommended", "checklist", "remediation"],
+        },
+    },
+    {
+        name: "massive-refactor",
+        hash: "",
+        promptPath: "00_setup/00-1_coach",
+        vars: {
+            objective: "Migrate the entire backend from Express to NestJS.",
+            tree: "src/app.ts\nsrc/routes/api.ts\nsrc/models/user.ts",
+            readme: "Express backend",
+            providers: "anthropic, google",
+            isInitialCoach: true,
+        },
+        criteria: {
+            independentTasks: false,
+            specificTasks: false,
+            requiredJsonFields: ["objective", "scope", "recommended", "checklist", "remediation"],
+        },
+    },
+];
+function hashCase(c) {
+    const key = `${c.promptPath}:${c.variant ?? "default"}:${JSON.stringify(c.vars)}`;
+    let h = 0;
+    for (let i = 0; i < key.length; i++) {
+        h = ((h << 5) - h + key.charCodeAt(i)) | 0;
+    }
+    return Math.abs(h).toString(36).slice(0, 8);
+}
+for (const c of COACH_CASES) {
+    c.hash = hashCase(c);
+}

package/dist/prompt-evolution/fixtures/steer-cases.d.ts

@@ -0,0 +1,2 @@
+import type { BenchmarkCase } from "../types.js";
+export declare const STEER_CASES: BenchmarkCase[];

package/dist/prompt-evolution/fixtures/steer-cases.js

@@ -0,0 +1,85 @@
+export const STEER_CASES = [
+    {
+        name: "idle-needs-verify",
+        hash: "",
+        promptPath: "30_wave/30-1_steer",
+        vars: {
+            objective: "Fix the pagination bug and ensure it works on mobile.",
+            status: "Pagination logic has been rewritten. Tests pass. We have not run the app to check mobile responsiveness yet.",
+            recentText: "Wave 2 finished: 1 feature agent refactored the logic.",
+            fastModel: "qwen",
+            workerModel: "sonnet",
+            longArchetypes: true,
+        },
+        criteria: {
+            independentTasks: true,
+            specificTasks: false,
+            requiredJsonFields: ["done", "reasoning", "statusUpdate", "estimatedSessionsRemaining", "tasks"],
+        },
+    },
+    {
+        name: "infinite-loop-stuck",
+        hash: "",
+        promptPath: "30_wave/30-1_steer",
+        vars: {
+            objective: "Migrate the database to PostgreSQL.",
+            status: "Agent keeps failing to connect to the database. It has tried 5 times with different credentials.",
+            recentText: "Wave 4 finished: Agent failed to run migrations due to 'Connection refused'.",
+            fastModel: "qwen",
+            workerModel: "sonnet",
+            longArchetypes: true,
+        },
+        criteria: {
+            independentTasks: true,
+            specificTasks: false,
+            requiredJsonFields: ["done", "reasoning", "statusUpdate", "estimatedSessionsRemaining", "tasks"],
+        },
+    },
+    {
+        name: "completed-all-requirements",
+        hash: "",
+        promptPath: "30_wave/30-1_steer",
+        vars: {
+            objective: "Add a dark mode toggle to the header.",
+            status: "Toggle component created, state management added, CSS updated. Verification agent confirmed the toggle works and persists across reloads.",
+            recentText: "Wave 3 finished: Verification agent reported full success.",
+            fastModel: "qwen",
+            workerModel: "sonnet",
+            longArchetypes: true,
+        },
+        criteria: {
+            independentTasks: true,
+            specificTasks: false,
+            requiredJsonFields: ["done", "reasoning", "statusUpdate", "estimatedSessionsRemaining", "tasks"],
+        },
+    },
+    {
+        name: "mid-feature-split",
+        hash: "",
+        promptPath: "30_wave/30-1_steer",
+        vars: {
+            objective: "Build a new analytics dashboard with 3 charts.",
+            status: "Database queries are written. We need to build the UI components and wire them up.",
+            recentText: "Wave 1 finished: Backend agent successfully added the SQL views.",
+            fastModel: "qwen",
+            workerModel: "sonnet",
+            longArchetypes: true,
+        },
+        criteria: {
+            independentTasks: true,
+            specificTasks: true,
+            requiredJsonFields: ["done", "reasoning", "statusUpdate", "estimatedSessionsRemaining", "tasks"],
+        },
+    }
+];
+function hashCase(c) {
+    const key = `${c.promptPath}:${c.variant ?? "default"}:${JSON.stringify(c.vars)}`;
+    let h = 0;
+    for (let i = 0; i < key.length; i++) {
+        h = ((h << 5) - h + key.charCodeAt(i)) | 0;
+    }
+    return Math.abs(h).toString(36).slice(0, 8);
+}
+for (const c of STEER_CASES) {
+    c.hash = hashCase(c);
+}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "claude-overnight",
-  "version": "1.60.1",
+  "version": "1.60.2",
   "description": "Overnight parallel coding agents in git worktrees, with a self-curating skill memory that improves while the run is going. Mix Claude Opus as planner, Kimi 2.6 or Cursor composer-2 as cheap fast worker, Gemini or Qwen for bulk implementation. Multi-wave autonomous loop that plans, executes, reviews, and steers itself until the objective is met. Crash-safe resume, rate-limit aware, usage cap preserves headroom for your interactive Claude Code.",
   "type": "module",
   "bin": {

package/plugins/claude-overnight/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "claude-overnight",
-  "version": "1.60.1",
+  "version": "1.60.2",
   "description": "Claude Code skill for understanding, installing, and inspecting claude-overnight runs: overnight parallel coding agents in git worktrees with a self-curating skill memory, multi-wave steering, three-layer review, and crash-safe resume. Mix Opus planner with Kimi 2.6, Cursor composer-2, Gemini, Qwen, or any Anthropic-compatible worker.",
   "author": {
     "name": "Francesco Fornace"