omnius 1.0.214 → 1.0.216

package/dist/index.js

@@ -24413,7 +24413,7 @@ var EXCLUDED, MAX_ENTRIES, ListDirectoryTool;
 var init_list_directory = __esm({
   "packages/execution/dist/tools/list-directory.js"() {
     "use strict";
-    EXCLUDED = /* @__PURE__ */ new Set(["node_modules", ".git"]);
+    EXCLUDED = /* @__PURE__ */ new Set(["node_modules", ".git", ".omnius"]);
     MAX_ENTRIES = 100;
     ListDirectoryTool = class {
       name = "list_directory";
@@ -289853,6 +289853,53 @@ function getTodoSessionId() {
     return envSession;
   return "default";
 }
+function normalizeIncomingTodos(args) {
+  const repairNotes = [];
+  const record = args;
+  if (Array.isArray(args)) {
+    repairNotes.push("coerced top-level array into {todos:[...]}");
+    return { todos: args, repairNotes, error: "" };
+  }
+  const direct = record["todos"];
+  if (Array.isArray(direct)) {
+    return { todos: direct, repairNotes, error: "" };
+  }
+  if (direct && typeof direct === "object") {
+    const nested = direct;
+    for (const key of ["todos", "items", "tasks", "checklist"]) {
+      if (Array.isArray(nested[key])) {
+        repairNotes.push(`coerced todos.${key} into todos array`);
+        return { todos: nested[key], repairNotes, error: "" };
+      }
+    }
+    if (typeof nested["content"] === "string") {
+      repairNotes.push("wrapped single todo object in todos array");
+      return { todos: [nested], repairNotes, error: "" };
+    }
+  }
+  for (const key of ["items", "tasks", "checklist", "todo_items"]) {
+    if (Array.isArray(record[key])) {
+      repairNotes.push(`coerced ${key} into todos array`);
+      return { todos: record[key], repairNotes, error: "" };
+    }
+  }
+  const single = record["todo"] ?? record["task"];
+  if (single && typeof single === "object" && !Array.isArray(single)) {
+    const obj = single;
+    if (typeof obj["content"] === "string") {
+      repairNotes.push("coerced single todo/task object into todos array");
+      return { todos: [obj], repairNotes, error: "" };
+    }
+  }
+  if (typeof single === "string" && single.trim()) {
+    repairNotes.push("coerced single todo/task string into todos array");
+    return { todos: [single.trim()], repairNotes, error: "" };
+  }
+  return {
+    repairNotes,
+    error: 'todos must be an array. Correct shape: todo_write({"todos":[{"content":"Inspect files","status":"in_progress"},{"content":"Make changes","status":"pending"}]})'
+  };
+}
 var _currentSessionId, TodoWriteTool, TodoReadTool;
 var init_todo_write = __esm({
   "packages/execution/dist/tools/todo-write.js"() {
@@ -289861,7 +289908,27 @@ var init_todo_write = __esm({
     _currentSessionId = "";
     TodoWriteTool = class {
       name = "todo_write";
-      description = "Update the session task checklist. To be used proactively and often to track progress and pending tasks. Make sure that at least one task is in_progress at all times. \n\n## When to use\n1. Complex multi-step tasks — when a task requires 3 or more distinct steps or actions\n2. When the user provides multiple tasks (numbered or comma-separated)\n3. After receiving new instructions — capture user requirements as todos immediately\n4. When you start a task — mark it in_progress BEFORE beginning work. Only ONE in_progress at a time\n5. After completing a task — mark it completed and add follow-up tasks you discovered\n\n## When NOT to use\n- Single, straightforward tasks (a trivial edit, a one-line fix)\n- Conversational or informational questions\n- Tasks completable in <3 trivial steps\n\n## Task states\n- pending: not started\n- in_progress: currently working on (exactly ONE at a time)\n- completed: fully done (tests pass, code works, goal met)\n- blocked: stuck on a dependency (include blocker text)\n\nMark tasks complete IMMEDIATELY after finishing — don't batch. Never mark completed if tests are failing or implementation is partial. The user watches this list in the chat UI in real time.";
+      description = `Update the session task checklist. To be used proactively and often to track progress and pending tasks. Make sure that at least one task is in_progress at all times. 
+
+## When to use
+1. Complex multi-step tasks — when a task requires 3 or more distinct steps or actions
+2. When the user provides multiple tasks (numbered or comma-separated)
+3. After receiving new instructions — capture user requirements as todos immediately
+4. When you start a task — mark it in_progress BEFORE beginning work. Only ONE in_progress at a time
+5. After completing a task — mark it completed and add follow-up tasks you discovered
+
+## When NOT to use
+- Single, straightforward tasks (a trivial edit, a one-line fix)
+- Conversational or informational questions
+- Tasks completable in <3 trivial steps
+
+## Task states
+- pending: not started
+- in_progress: currently working on (exactly ONE at a time)
+- completed: fully done (tests pass, code works, goal met)
+- blocked: stuck on a dependency (include blocker text)
+
+Mark tasks complete IMMEDIATELY after finishing — don't batch. Never mark completed if tests are failing or implementation is partial. The user watches this list in the chat UI in real time. Canonical call shape: todo_write({"todos":[{"content":"Inspect files","status":"in_progress"},{"content":"Make changes","status":"pending"},{"content":"Verify results","status":"pending"}]})`;
       parameters = {
         type: "object",
         required: ["todos"],
@@ -289902,48 +289969,62 @@ var init_todo_write = __esm({
       async execute(args) {
         const start2 = performance.now();
         try {
-          const incomingRaw = args["todos"];
-          if (!Array.isArray(incomingRaw)) {
+          const normalized = normalizeIncomingTodos(args);
+          if (!normalized.todos) {
             return {
               success: false,
               output: "",
-              error: "todos must be an array",
+              error: normalized.error,
               durationMs: performance.now() - start2
             };
           }
           const incoming = [];
-          for (const raw of incomingRaw) {
+          const repairNotes = [...normalized.repairNotes];
+          for (let index = 0; index < normalized.todos.length; index++) {
+            const raw = normalized.todos[index];
             if (!raw || typeof raw !== "object") {
+              if (typeof raw === "string" && raw.trim()) {
+                incoming.push({
+                  content: raw.trim(),
+                  status: index === 0 ? "in_progress" : "pending"
+                });
+                repairNotes.push("coerced string todo item into {content,status}");
+                continue;
+              }
               return {
                 success: false,
                 output: "",
-                error: "each todo must be an object with content+status",
+                error: 'each todo must be an object with content+status. Correct shape: {"todos":[{"content":"...","status":"in_progress"}]}',
                 durationMs: performance.now() - start2
               };
             }
             const entry = raw;
             const content = entry["content"];
             const status = entry["status"];
-            if (typeof content !== "string" || typeof status !== "string") {
+            if (typeof content !== "string") {
               return {
                 success: false,
                 output: "",
-                error: "todo must have string content and string status",
+                error: 'todo must have string content. Correct shape: {"todos":[{"content":"...","status":"in_progress"}]}',
                 durationMs: performance.now() - start2
               };
             }
-            if (!["pending", "in_progress", "completed", "blocked"].includes(status)) {
+            const resolvedStatus = typeof status === "string" ? status : index === 0 ? "in_progress" : "pending";
+            if (typeof status !== "string") {
+              repairNotes.push("defaulted missing todo status to in_progress/pending");
+            }
+            if (!["pending", "in_progress", "completed", "blocked"].includes(resolvedStatus)) {
               return {
                 success: false,
                 output: "",
-                error: `invalid status: ${status}`,
+                error: `invalid status: ${resolvedStatus}`,
                 durationMs: performance.now() - start2
               };
             }
             incoming.push({
               id: typeof entry["id"] === "string" ? entry["id"] : void 0,
               content,
-              status,
+              status: resolvedStatus,
               parentId: typeof entry["parentId"] === "string" ? entry["parentId"] : void 0,
               blocker: typeof entry["blocker"] === "string" ? entry["blocker"] : void 0,
               // REG-37: verification-aware planning
@@ -289986,6 +290067,16 @@ var init_todo_write = __esm({
             newTodos: result.newTodos,
             verificationNudgeNeeded
           };
+          if (repairNotes.length > 0) {
+            payload["inputRepair"] = Array.from(new Set(repairNotes));
+            payload["canonicalShape"] = {
+              todos: [
+                { content: "Inspect files", status: "in_progress" },
+                { content: "Make changes", status: "pending" },
+                { content: "Verify results", status: "pending" }
+              ]
+            };
+          }
           if (verificationNudgeNeeded) {
             payload["nudge"] = "You just closed 3+ todos without scheduling a verification step. Add a 'Verify the changes work' item and spawn a verification agent before declaring task_complete.";
           }
@@ -564459,10 +564550,12 @@ ${_staleSamples.join("\n")}` : ``,
           const turnTier = this.options.modelTier ?? "large";
           if (turn === 0 && !this.options.disableTodoPlanningNudges && (turnTier === "small" || turnTier === "medium")) {
             const goal = this._taskState.goal || "";
-            const wordCount2 = goal.split(/\s+/).length;
-            const hasMultipleActions = /\band\b.*\band\b|then.*then|also.*also/i.test(goal);
-            const hasMultipleFiles = /files?.*files?|\.ts.*\.ts|create.*write|modify.*create/i.test(goal);
-            const isComplex = wordCount2 > 40 || hasMultipleActions || hasMultipleFiles;
+            const substantiveGoal = goal.replace(/\b(?:then\s+)?call\s+task_complete\b[^.?!;]*/gi, "").replace(/\b(?:observe|report|summarize|finish|complete)\b[^.?!;]*/gi, "");
+            const wordCount2 = substantiveGoal.split(/\s+/).filter(Boolean).length;
+            const hasMultipleActions = /\band\b.*\band\b|then.*then|also.*also/i.test(substantiveGoal);
+            const hasMultipleFiles = /files?.*files?|\.ts.*\.ts|create.*write|modify.*create/i.test(substantiveGoal);
+            const explicitSingleTool = /\b(exactly once|single tool|one tool|one tool call)\b/i.test(goal) || /\b(call|use)\s+(?:list_directory|file_read|grep_search|find_files|shell|web_search|web_fetch)\b/i.test(goal) && !/\b(edit|write|modify|create|fix|implement|patch|test|build|install|refactor)\b/i.test(substantiveGoal);
+            const isComplex = !explicitSingleTool && (wordCount2 > 40 || hasMultipleActions || hasMultipleFiles);
             if (isComplex) {
               messages2.push({
                 role: "user",
@@ -564471,6 +564564,7 @@ ${_staleSamples.join("\n")}` : ``,
 MANDATORY FIRST ACTION: Call todo_write NOW with the complete plan.
 Each todo item is { content: "what to do", status: "pending" | "in_progress" | "completed" | "blocked" }.
 Mark item 1 as in_progress, the rest as pending.
+Only count substantive work phases. Do NOT count observing a tool result, reporting findings, or calling task_complete as todo phases.
 Example: todo_write({todos: [{content: "read source files", status: "in_progress"}, {content: "make changes", status: "pending"}, {content: "run tests", status: "pending"}]})
 
 After EACH phase finishes, call todo_write AGAIN with item N marked completed and item N+1 marked in_progress.
@@ -564566,7 +564660,7 @@ ${top.map((t2) => `- ${t2.name}: ${t2.desc}`).join("\n")}`);
             const isReadTask = /\bread\b|\bshow\b|\btell me\b|\bwhat is\b/i.test(taskGoal);
             const hints = [];
             if (isSimpleTask) {
-              hints.push("This is a simple task — if it needs only ONE tool call, skip todo_write and call the tool directly. If it needs 2+ steps, use todo_write to plan.");
+              hints.push("This is a simple task — if it needs only ONE substantive tool call, skip todo_write and call the tool directly, then task_complete. Do not count reporting, observing output, or task_complete as planning steps. If it needs 2+ substantive work steps, use todo_write to plan.");
             }
             if (isSearchTask) {
               hints.push("SEARCH STRATEGY: Use grep_search to find what you need FIRST, THEN file_read only the specific file and lines. Do NOT read entire files hoping to find something.");
@@ -578422,6 +578516,70 @@ var init_generative_progress = __esm({
   }
 });
 
+// packages/cli/src/tui/tool-adapter.ts
+function mapExecutionToolResult(result) {
+  return {
+    success: result.success,
+    output: result.output,
+    error: result.error,
+    llmContent: result.llmContent,
+    mutated: result.mutated,
+    mutatedFiles: result.mutatedFiles,
+    diff: result.diff,
+    dryRun: result.dryRun,
+    noop: result.noop,
+    partial: result.partial,
+    beforeHash: result.beforeHash,
+    afterHash: result.afterHash
+  };
+}
+function adaptExecutionTool(tool, options2 = {}) {
+  const progressTool = tool;
+  if (generationKindForToolName(tool.name) && typeof progressTool.setProgressCallback === "function") {
+    progressTool.setProgressCallback((event) => {
+      options2.onProgress?.(tool.name, event);
+    });
+  }
+  const adapted = {
+    name: tool.name,
+    aliases: tool.aliases,
+    description: tool.description,
+    parameters: tool.parameters,
+    inputSchema: tool.inputSchema,
+    maxResultSizeChars: tool.maxResultSizeChars,
+    async execute(args) {
+      const invoke = () => tool.execute(args);
+      const result = options2.execute ? await options2.execute(tool, args, invoke) : await invoke();
+      return mapExecutionToolResult(result);
+    }
+  };
+  if (typeof tool.prompt === "function") {
+    adapted.prompt = (context2) => tool.prompt(context2);
+  }
+  if (typeof tool.executeStream === "function") {
+    adapted.executeStream = async function* (args) {
+      const result = yield* tool.executeStream(args);
+      return mapExecutionToolResult(result);
+    };
+  }
+  if (typeof tool.validateInput === "function") {
+    adapted.validateInput = (args, context2) => tool.validateInput(args, context2);
+  }
+  if (typeof tool.isConcurrencySafe === "function") {
+    adapted.isConcurrencySafe = (args) => tool.isConcurrencySafe(args);
+  }
+  if (typeof tool.isReadOnly === "function") {
+    adapted.isReadOnly = (args) => tool.isReadOnly(args);
+  }
+  return adapted;
+}
+var init_tool_adapter = __esm({
+  "packages/cli/src/tui/tool-adapter.ts"() {
+    "use strict";
+    init_generative_progress();
+  }
+});
+
 // packages/cli/src/tui/runtime-verification.ts
 import { execFileSync as execFileSync6 } from "node:child_process";
 import { existsSync as existsSync92, readFileSync as readFileSync74, readdirSync as readdirSync29 } from "node:fs";
@@ -592024,15 +592182,7 @@ var init_p2p = __esm({
 import { EventEmitter as EventEmitter11 } from "node:events";
 import crypto13 from "node:crypto";
 function adaptTool(tool) {
-  return {
-    name: tool.name,
-    description: tool.description,
-    parameters: tool.parameters,
-    async execute(args) {
-      const result = await tool.execute(args);
-      return { success: result.success, output: result.output, error: result.error };
-    }
-  };
+  return adaptExecutionTool(tool);
 }
 function getActivityFeed() {
   if (!_globalFeed) _globalFeed = new ActivityFeed();
@@ -592047,6 +592197,7 @@ var init_call_agent = __esm({
     "use strict";
     init_dist8();
     init_dist6();
+    init_tool_adapter();
     ActivityFeed = class {
       entries = [];
       maxEntries = 100;
@@ -632130,15 +632281,7 @@ function computeSparsity(entries) {
   return Math.max(0, Math.min(1, 1 - avgOverlap));
 }
 function adaptTool2(tool) {
-  return {
-    name: tool.name,
-    description: tool.description,
-    parameters: tool.parameters,
-    async execute(args) {
-      const result = await tool.execute(args);
-      return { success: result.success, output: result.output, error: result.error };
-    }
-  };
+  return adaptExecutionTool(tool);
 }
 var SNREngine;
 var init_snr_engine = __esm({
@@ -632148,6 +632291,7 @@ var init_snr_engine = __esm({
     init_dist6();
     init_project_context();
     init_render();
+    init_tool_adapter();
     SNREngine = class {
       constructor(config, repoRoot) {
         this.config = config;
@@ -632608,15 +632752,7 @@ ${sections.join("\n\n")}`;
   }
 }
 function adaptTool3(tool) {
-  return {
-    name: tool.name,
-    description: tool.description,
-    parameters: tool.parameters,
-    async execute(args) {
-      const result = await tool.execute(args);
-      return { success: result.success, output: result.output, error: result.error };
-    }
-  };
+  return adaptExecutionTool(tool);
 }
 function buildDreamPrompt(mode, stage, cycleNum, totalCycles, previousFindings, dreamsDir) {
   const modeDesc = mode === "lucid" ? "LUCID DREAM MODE: You have full implementation capability. After ideation, you will implement, test, and evaluate changes." : mode === "deep" ? "DEEP DREAM MODE: Explore deeply with multiple expansion/contraction cycles. All proposals go in .omnius/dreams/." : "DREAM MODE: Creative exploration only. All output must be written to .omnius/dreams/ directory using file_write.";
@@ -632758,6 +632894,7 @@ var init_dream_engine = __esm({
     init_setup();
     init_render();
     init_promptLoader3();
+    init_tool_adapter();
     _dreamWriteContent = null;
     SWARM_ROLE_CONFIG = {
       researcher: { maxTurns: 25, temperature: 0.4 },
@@ -634546,15 +634683,7 @@ Reflect on what went well and what could improve.`;
   }
 }
 function adaptTool4(tool) {
-  return {
-    name: tool.name,
-    description: tool.description,
-    parameters: tool.parameters,
-    async execute(args) {
-      const result = await tool.execute(args);
-      return { success: result.success, output: result.output, error: result.error };
-    }
-  };
+  return adaptExecutionTool(tool);
 }
 function renderDMNCycleStart(cycleNum, deliberation = false) {
   process.stdout.write(`
@@ -634622,6 +634751,7 @@ var init_dmn_engine = __esm({
     init_project_context();
     init_render();
     init_promptLoader3();
+    init_tool_adapter();
     DMNEngine = class {
       constructor(config, repoRoot) {
         this.config = config;
@@ -642252,25 +642382,17 @@ function normalizeTelegramCallbackQuery(update2) {
   };
 }
 function adaptTool5(tool, todoSessionId, progress) {
-  const progressTool = tool;
-  if (generationKindForToolName(tool.name) && typeof progressTool.setProgressCallback === "function") {
-    progressTool.setProgressCallback((event) => {
-      progress?.onProgress(tool.name, event);
-    });
-  }
-  return {
-    name: tool.name,
-    description: tool.description,
-    parameters: tool.parameters,
-    async execute(args) {
+  return adaptExecutionTool(tool, {
+    onProgress: (toolName, event) => progress?.onProgress(toolName, event),
+    execute: async (_tool, args, invoke) => {
       const previousTodoSession = todoSessionId ? getTodoSessionId() : "";
       if (todoSessionId && (tool.name === "todo_write" || tool.name === "todo_read")) {
         setTodoSessionId(todoSessionId);
       }
       try {
-        const result = await tool.execute(args);
+        const result = await invoke();
         progress?.complete(tool.name, result);
-        return { success: result.success, output: result.output, error: result.error, llmContent: result.llmContent };
+        return result;
       } catch (err) {
         progress?.complete(tool.name, {
           success: false,
@@ -642284,7 +642406,7 @@ function adaptTool5(tool, todoSessionId, progress) {
         }
       }
     }
-  };
+  });
 }
 function telegramBotAccessSettingsFromApi(settings) {
   return {
@@ -642449,6 +642571,7 @@ var init_telegram_bridge = __esm({
     init_voice_soul();
     init_telegram_creative_tools();
     init_generative_progress();
+    init_tool_adapter();
     init_omnius_directory();
     init_stimulation();
     init_pid_controller();
@@ -681209,42 +681332,9 @@ function getVersion4() {
   return "0.0.0";
 }
 function adaptTool6(tool) {
-  const progressTool = tool;
-  if (generationKindForToolName(tool.name) && typeof progressTool.setProgressCallback === "function") {
-    progressTool.setProgressCallback((event) => {
-      _generativeProgressSink?.(tool.name, event);
-    });
-  }
-  return {
-    name: tool.name,
-    aliases: tool.aliases,
-    description: tool.description,
-    parameters: tool.parameters,
-    inputSchema: tool.inputSchema,
-    prompt: tool.prompt,
-    executeStream: tool.executeStream,
-    validateInput: tool.validateInput,
-    isConcurrencySafe: tool.isConcurrencySafe,
-    isReadOnly: tool.isReadOnly,
-    maxResultSizeChars: tool.maxResultSizeChars,
-    async execute(args) {
-      const result = await tool.execute(args);
-      return {
-        success: result.success,
-        output: result.output,
-        error: result.error,
-        llmContent: result.llmContent,
-        mutated: result.mutated,
-        mutatedFiles: result.mutatedFiles,
-        diff: result.diff,
-        dryRun: result.dryRun,
-        noop: result.noop,
-        partial: result.partial,
-        beforeHash: result.beforeHash,
-        afterHash: result.afterHash
-      };
-    }
-  };
+  return adaptExecutionTool(tool, {
+    onProgress: (toolName, event) => _generativeProgressSink?.(toolName, event)
+  });
 }
 function createTuiReminderOptions(allowActionDelivery = true) {
   const sessionId = process.env["OMNIUS_SESSION_ID"] || "terminal";
@@ -690440,6 +690530,7 @@ var init_interactive = __esm({
     init_dist8();
     init_dist6();
     init_generative_progress();
+    init_tool_adapter();
     init_runtime_verification();
     init_dist();
     init_listen();
@@ -691679,6 +691770,7 @@ function parseCliArgs(argv) {
       local: { type: "boolean", short: "l" },
       port: { type: "string" },
       suite: { type: "string" },
+      live: { type: "boolean" },
       json: { type: "boolean", short: "j" },
       background: { type: "boolean" },
       help: { type: "boolean", short: "h" },
@@ -691738,6 +691830,7 @@ function parseCliArgs(argv) {
         break;
       case "eval":
         result.evalSuite = typeof values.suite === "string" ? values.suite : void 0;
+        result.evalLive = values.live === true;
         break;
       default:
         break;
@@ -691776,6 +691869,7 @@ Flags:
       --max-retries <n>      Max retries per model request
       --timeout-ms <ms>      Overall task timeout
       --suite <name>         Eval suite: basic (default) or full
+      --live                 Run eval against configured backend instead of FakeBackend
       --port <n>             Server port (serve command, vLLM only, default: 8000)
   -h, --help                 Show this help
   -V, --version              Show version
@@ -691801,6 +691895,7 @@ Examples:
   omnius serve
   omnius serve --backend vllm --port 9000
   omnius eval --suite full --verbose
+  omnius eval --suite basic --live --backend ollama --model qwen3.5:9b
   omnius config set model qwen3.5:122b
 `.trim();
   process.stdout.write(text + "\n");
@@ -691936,7 +692031,8 @@ async function main() {
           {
             suite: parsed.evalSuite,
             repoPath: parsed.repoPath,
-            verbose: parsed.verbose
+            verbose: parsed.verbose,
+            live: parsed.evalLive
           },
           config
         );

package/npm-shrinkwrap.json

@@ -1,12 +1,12 @@
 {
   "name": "omnius",
-  "version": "1.0.214",
+  "version": "1.0.216",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "omnius",
-      "version": "1.0.214",
+      "version": "1.0.216",
       "bundleDependencies": [
         "image-to-ascii"
       ],

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "omnius",
-  "version": "1.0.214",
+  "version": "1.0.216",
   "description": "AI coding agent powered by open-source models (Ollama/vLLM) — interactive TUI with agentic tool-calling loop",
   "type": "module",
   "main": "./dist/index.js",

package/prompts/agentic/system-large.md

@@ -51,7 +51,7 @@ If you anticipate a large result before calling a tool, prefer narrow flags firs
 - list_directory: List files in a directory with types and sizes
 - web_search: Search the web for documentation or solutions
 - web_fetch: Fetch a web page and extract text content (for docs, MDN, w3schools.com, etc.)
-- todo_write / todo_read: Visible task checklist for the user. For ANY multi-step task with 3+ logical phases, your FIRST tool call must be todo_write declaring the entire plan as an array of items with status pending|in_progress|completed|blocked. After each phase completes, call todo_write again with item N marked completed and item N+1 marked in_progress. The user watches this checklist update live in the chat UI — it is your primary planning surface for long-horizon work and the user can see at a glance whether you are making progress or stuck. Use todo_write for any task naturally containing 3+ phases (build/test/ship, scrape/parse/store, plan/draft/edit, explore/refactor/verify, etc.). Do NOT use it for trivial single-step questions. Each todo accepts two OPTIONAL fields you should USE whenever the todo has objective completion criteria: `verifyCommand` (a shell command that PROVES the todo is complete — typecheck/test/build invocations etc.) and `declaredArtifacts` (a list of file paths this todo will produce). The orchestrator auto-checks both at completion-claim time; missing/unverified completions are rejected with a specific gap critique. **Worked example — emit todos in this exact shape:** `todo_write({"todos":[{"id":"p1","content":"Implement cache module","status":"in_progress","verifyCommand":"<your test command>","declaredArtifacts":["src/lib/cache.ts","tests/cache.test"]},{"id":"p2","content":"Make build pass","status":"pending","verifyCommand":"<your build command>"}]})`. Substitute placeholder strings with commands native to YOUR stack.
+- todo_write / todo_read: Visible task checklist for the user. For ANY multi-step task with 3+ substantive work phases, your FIRST tool call must be todo_write declaring the entire plan as an array of items with status pending|in_progress|completed|blocked. After each phase completes, call todo_write again with item N marked completed and item N+1 marked in_progress. Do NOT count observing a tool result, reporting findings, or task_complete as phases. The user watches this checklist update live in the chat UI — it is your primary planning surface for long-horizon work and the user can see at a glance whether you are making progress or stuck. Use todo_write for any task naturally containing 3+ real work phases (build/test/ship, scrape/parse/store, plan/draft/edit, explore/refactor/verify, etc.). Do NOT use it for trivial single-step questions. Each todo accepts two OPTIONAL fields you should USE whenever the todo has objective completion criteria: `verifyCommand` (a shell command that PROVES the todo is complete — typecheck/test/build invocations etc.) and `declaredArtifacts` (a list of file paths this todo will produce). The orchestrator auto-checks both at completion-claim time; missing/unverified completions are rejected with a specific gap critique. **Worked example — emit todos in this exact shape:** `todo_write({"todos":[{"id":"p1","content":"Implement cache module","status":"in_progress","verifyCommand":"<your test command>","declaredArtifacts":["src/lib/cache.ts","tests/cache.test"]},{"id":"p2","content":"Make build pass","status":"pending","verifyCommand":"<your build command>"}]})`. Substitute placeholder strings with commands native to YOUR stack.
 
 ## Web Tool Selection
 
@@ -182,7 +182,7 @@ When you discover image files (png, jpg, gif, svg, webp, bmp) during codebase ex
 
 ## Workflow
 
-0. **PLAN AT THE TOP** — for any task with 3+ logical phases, your VERY FIRST tool call must be `todo_write` with a complete checklist (each item: `{content, status}`). Mark item 1 as `in_progress`, the rest as `pending`. The user watches this checklist update live in the chat UI as you work, so they always know what step you're on. After each phase, call todo_write again to mark the finished item `completed` and the next one `in_progress`.
+0. **PLAN AT THE TOP** — for any task with 3+ substantive work phases, your VERY FIRST tool call must be `todo_write` with a complete checklist (each item: `{content, status}`). Mark item 1 as `in_progress`, the rest as `pending`. Do not count observing output, reporting findings, or task_complete as phases. The user watches this checklist update live in the chat UI as you work, so they always know what step you're on. After each phase, call todo_write again to mark the finished item `completed` and the next one `in_progress`.
 1. EXPLORE: Use find_files and grep_search to locate relevant code. Read specific files.
 2. PLAN: Determine what changes are needed based on the code you've read.
 3. IMPLEMENT: Make changes using file_edit (preferred) or file_write for new files.

package/prompts/agentic/system-medium.md

@@ -11,7 +11,7 @@ You operate in two modes based on what the user needs:
 **TASK MODE** — coding tasks, file operations, technical directives:
 - Call tools iteratively until complete. NEVER write code blocks as text — only tool calls execute.
 - If you need to read a file, call file_read. If you need to run a command, call shell.
-- **MANDATORY: For ANY task that will take 3 or more tool calls, your VERY FIRST tool call MUST be `todo_write` declaring the complete plan.** Items have `{content, status}` where status is one of pending|in_progress|completed|blocked. Mark item 1 in_progress, the rest pending. Then re-call todo_write after each phase finishes to mark item N completed and N+1 in_progress. The user watches this checklist update live in the chat UI — without it they can't see your plan or track your progress.
+- **MANDATORY: For ANY task that will take 3 or more substantive work tool calls, your VERY FIRST tool call MUST be `todo_write` declaring the complete plan.** Items have `{content, status}` where status is one of pending|in_progress|completed|blocked. Mark item 1 in_progress, the rest pending. Then re-call todo_write after each phase finishes to mark item N completed and N+1 in_progress. Do NOT count observing tool output, reporting findings, or task_complete as work phases. For one-tool tasks, call the tool directly and then task_complete. The user watches this checklist update live in the chat UI — without it they can't see your plan or track your progress.
 
 ## Instruction Hierarchy
 
@@ -41,7 +41,7 @@ Tool results over ~100KB are NOT truncated. The orchestrator saves the full payl
 - list_directory: List files in a directory
 - web_search: Search the web
 - web_fetch: Fetch a web page's text
-- todo_write / todo_read: Visible task checklist for the user. For ANY multi-step task with 3+ logical steps, start by calling todo_write to declare your plan, then re-call todo_write as each step transitions (mark item N "completed" + N+1 "in_progress"). The user sees this list update live in the UI — it is your primary planning surface for long-horizon work. Use it whenever the task naturally has 3+ phases (build/refactor/test/ship, scrape/parse/store/report, plan/draft/edit/publish, etc.).
+- todo_write / todo_read: Visible task checklist for the user. For ANY multi-step task with 3+ substantive work steps, start by calling todo_write to declare your plan, then re-call todo_write as each step transitions (mark item N "completed" + N+1 "in_progress"). The user sees this list update live in the UI — it is your primary planning surface for long-horizon work. Use it whenever the task naturally has 3+ real work phases (build/refactor/test/ship, scrape/parse/store/report, plan/draft/edit/publish, etc.). Skip it for a single tool action followed only by reporting and task_complete.
 
   Each todo accepts two OPTIONAL fields you should USE whenever the todo has objective completion criteria:
 
@@ -105,8 +105,8 @@ Launch ALL sub_agent calls in ONE response. This saves your context window for o
 
 ## Workflow
 
-For tasks requiring 3+ tool calls — plan before acting:
-1. LIST all steps needed before your first tool call. **For 3+ step tasks, your FIRST tool call must be `todo_write` declaring the full plan with item 1 set to status:"in_progress" and the rest "pending".** Then call todo_write again as each step finishes to mark items "completed" and the next one "in_progress". The user watches this list update live in the chat UI.
+For tasks requiring 3+ substantive work tool calls — plan before acting:
+1. LIST all real work steps needed before your first tool call. **For 3+ substantive-step tasks, your FIRST tool call must be `todo_write` declaring the full plan with item 1 set to status:"in_progress" and the rest "pending".** Do not count reporting, observing output, or task_complete as steps. Then call todo_write again as each step finishes to mark items "completed" and the next one "in_progress". The user watches this list update live in the chat UI.
 2. If task mentions 3+ independent modules/files: delegate each to a sub_agent (saves context)
 3. EXPLORE: Use find_files, grep_search, file_explore to understand the codebase
    - For large files (200+ lines): use file_explore(strategy='overview') then search/chunk — NEVER read entire file

package/prompts/agentic/system-small.md

@@ -34,7 +34,7 @@ File edits: Use file_write/file_edit/file_patch/batch_edit for project files, no
 
 Tool choice: Use file/search/code-graph tools for repository discovery, web_fetch/web_download/browser_action for web work, and repl_exec for multi-step data processing. Use shell when the command itself is the verifier or work product: tests, builds, package managers, git, system operations, and small native scripts. Do not hide diagnostics inside opaque shell blobs or `|| true`. Use background_run for long commands and poll with task_status/task_output.
 
-todo_write: visible task checklist for the user. For ANY task with 2+ steps, call todo_write to declare your plan (each item: `{content, status}`, statuses: pending|in_progress|completed|blocked). Update status as you complete each step. Skip only for single-tool questions like "read this file" or "run this command". Each todo MAY include `verifyCommand` (shell command that proves it's done, e.g. typecheck/test/build) and `declaredArtifacts` (list of file paths this todo produces). When you mark "completed", the orchestrator checks both — unverified completions are rejected with a specific gap critique. **Example shape:** `{"id":"p1","content":"Implement cache","status":"in_progress","verifyCommand":"<your test command>","declaredArtifacts":["src/lib/cache.ts"]}`. Substitute placeholders with commands native to YOUR stack.
+todo_write: visible task checklist for the user. Use it for substantive multi-step work, not ceremony. For tasks with 2+ substantive work steps, call todo_write to declare your plan (each item: `{content, status}`, statuses: pending|in_progress|completed|blocked). Update status as you complete each step. Skip single-tool questions like "read this file", "list this directory", or "run this command", even if you will report findings and call task_complete afterward. Do NOT count observing a tool result, reporting findings, or task_complete as todo steps. Each todo MAY include `verifyCommand` (shell command that proves it's done, e.g. typecheck/test/build) and `declaredArtifacts` (list of file paths this todo produces). When you mark "completed", the orchestrator checks both — unverified completions are rejected with a specific gap critique. **Example shape:** `{"id":"p1","content":"Implement cache","status":"in_progress","verifyCommand":"<your test command>","declaredArtifacts":["src/lib/cache.ts"]}`. Substitute placeholders with commands native to YOUR stack.
 
 Web: web_search finds URLs, web_fetch reads them. For JS pages use web_crawl, for clicking/login use browser_action.
 
@@ -100,7 +100,7 @@ Creating new files — WRITE FIRST, refine later:
 - After writing: fill in each method, test after each one.
 - A bad first draft you can fix is better than no draft at all.
 
-Complex tasks (5+ steps) — DECOMPOSE before acting:
+Complex tasks (5+ substantive work steps) — DECOMPOSE before acting:
 1. Call todo_write with the checklist. Mark item 1 "in_progress".
 2. Execute ONE STEP AT A TIME. After each, update todo_write status.
 3. After each file edit, VERIFY: file_read or shell test.