mixdog 0.9.151 → 0.9.153

package/LICENSES/browser-import-NOTICE.txt

@@ -0,0 +1,11 @@
+Chrome password import sidecar
+
+The Windows password importer packaged beside Mixdog is a separate
+GPL-3.0-only process derived from:
+
+https://github.com/bitwarden/clients
+commit 6e2c2151f215df69b7cf75b43f189b2cba8b6b5e
+
+Mixdog's wrapper and reproducible build instructions are available under
+native/mixdog-browser-import/. The packaged native-tools directory also carries
+the complete LICENSE_GPL.txt copied from that pinned source.

package/NOTICE.md

@@ -15,6 +15,8 @@ Full license texts live in `LICENSES/`:
   compiled into the native binaries.
 - `LICENSES/editor-assets-NOTICE.txt` — the editor icon and language-data
   notice referenced by the generated files.
+- `LICENSES/browser-import-NOTICE.txt` — source and license boundary for the
+  optional GPL Chrome password import sidecar.
 
 ## MIT
 
@@ -105,3 +107,12 @@ upstream NOTICE file is preserved as `LICENSES/codex-NOTICE.txt`.
 The Apache-2.0 terms require this notice to travel with any redistribution of
 the derived files. Full license text: `LICENSES/Apache-2.0.txt`
 (<https://www.apache.org/licenses/LICENSE-2.0>).
+
+## GPL-3.0-only
+
+### Chrome password import sidecar
+
+The Windows password importer is a separate process built from the
+pinned source and Mixdog wrapper recorded in
+`LICENSES/browser-import-NOTICE.txt`. It is not linked into the MIT desktop
+application. Its complete GPL text ships beside the two native executables.

package/README.md

@@ -1,20 +1,33 @@
-# mixdog
+# Mixdog
 
 [![npm](https://img.shields.io/npm/v/mixdog)](https://www.npmjs.com/package/mixdog)
 ![node](https://img.shields.io/badge/node-%3E%3D22-brightgreen)
 ![license](https://img.shields.io/badge/license-MIT-blue)
 
-Mixdog is a standalone coding agent for orchestrated, multi-provider workflows.
-Use it from a full-screen terminal UI or the Windows desktop app.
+## Better results. Less cost. More work.
+
+Mixdog is an efficiency-first AI coding harness designed to deliver equal or
+better performance with less context, time, and cost—so you can complete more
+work within the same API budget or subscription quota.
+
+With simple setup and an intuitive UX, Mixdog makes powerful orchestration,
+parallel tasks, and seamless work across terminal, desktop, and web accessible
+to everyone—from beginners to experts.
+
+**The easiest way to get more out of every coding model.**
 
 ## Get started
 
-### Windows desktop
+### Desktop
 
-[Download Mixdog Desktop for Windows (x64)](https://github.com/tribgames/mixdog/releases/latest/download/mixdog-desktop-win-x64.exe)
+<p align="center">
+  <a href="https://github.com/tribgames/mixdog/releases/latest/download/mixdog-desktop-win-x64.exe">
+    <img src="https://img.shields.io/badge/Download_for_Windows_x64-0078D4?style=for-the-badge&logo=windows11&logoColor=white" alt="Download Mixdog for Windows x64" height="56">
+  </a>
+</p>
 
-The desktop package is currently unsigned, so Windows SmartScreen may show a
-security warning during installation.
+The Windows installer is currently unsigned, so Windows SmartScreen may show a
+security warning.
 
 ### CLI
 
@@ -28,28 +41,68 @@ mixdog
 First run guides you through provider authentication, model selection, and
 workflow setup.
 
+## Benchmarks
+
+Terminal-Bench 2.1 — same model, same 89 tasks, same official verifier, with
+only the harness changed. Against the native CLI of each model family, Mixdog
+scores higher while spending less to get there.
+
+### GPT-5.6 Sol xhigh — Mixdog vs Codex CLI
+
+![Terminal-Bench 2.1: Mixdog with GPT-5.6 Sol xhigh versus Codex CLI](https://raw.githubusercontent.com/tribgames/mixdog/main/benchmarks/terminal-bench-2.1/tb21-sol-vs-codex.svg)
+
+- **86.5%** (385/445) vs Codex CLI's **84.3%** (75/89)
+- **42%** lower priced cost — $0.641 vs $1.096 per trial
+- **45%** smaller median final context — 18.5k vs 33.5k tokens
+- **1.11×** faster — 339s vs 378s per trial
+
+### Claude Opus 5 — Mixdog vs Claude Code
+
+![Terminal-Bench 2.1: Mixdog with Claude Opus 5 versus Claude Code](https://raw.githubusercontent.com/tribgames/mixdog/main/benchmarks/terminal-bench-2.1/tb21-opus-vs-claude-code.svg)
+
+- **79/89** vs Claude Code's **77/89**
+- **19%** lower priced cost — $104.29 vs $129.21 per run
+- **28%** smaller median final context — 27.6k vs 38.2k tokens
+- **1.15×** faster
+
+Every run uses the official Harbor verifier, fast mode off, a 272k context
+window, and zero retries. The Mixdog Sol run follows the protocol the official
+Terminal-Bench leaderboard requires — all 89 tasks repeated five times (`k=5`,
+445 trials); the Codex CLI baseline and both Opus-side runs are single passes
+(`k=1`, 89 trials each).
+
+The leaderboard is not accepting community submissions, so every run here ships
+its raw artifacts instead — Harbor verdicts, official verifier output, pinned
+task checksums, and the usage snapshots behind every cost figure — alongside
+the harness, presets, and metric scripts that recompute each number above:
+[`benchmarks/terminal-bench-2.1/`](benchmarks/terminal-bench-2.1/).
+
 ## Highlights
 
 - **Multi-provider routing** — assign different providers and models by role.
+- **Shared live sessions** — move between the TUI, desktop windows, and paired
+  browsers without starting a second copy of the session.
 - **Efficient context** — cache-aware prompts, compaction, resumable sessions,
   and focused repo-native tools.
 - **Complete coding surface** — read, search, edit, test, review, web search,
   MCP, skills, hooks, and plugins.
 - **Local memory** — semantic and lexical recall with project-scoped context
   and multilingual retrieval.
-- **Remote workflows** — optional web relay, Discord, Telegram, voice, and cron
-  schedules.
-- **Windows desktop app** — agent panes, Monaco editor, git, terminal, file
-  explorer, Studio, automation, and settings in one workbench.
+- **Encrypted remote access** — pair the installable web app with Desktop and
+  use Mixdog from a browser or phone over authenticated E2EE.
+- **Desktop coding app** — agent panes, Monaco editor, Git, terminals, file
+  explorer, Studio, automation, voice input, and settings in one app.
 
 ## Providers
 
 Mixdog supports subscription OAuth and API-key routes, including:
 
-- Anthropic and Claude accounts
-- OpenAI and ChatGPT/Codex accounts
-- Google Gemini
-- xAI Grok
+- Anthropic API keys and Claude account OAuth
+- OpenAI API keys and ChatGPT/Codex account OAuth
+- Google Gemini and Antigravity OAuth
+- xAI API keys and Grok account OAuth
+- OpenRouter API keys and its unified model catalog
+- Experimental Cursor account OAuth
 - DeepSeek and OpenCode Go
 - OpenAI-compatible APIs
 - Ollama and LM Studio
@@ -107,40 +160,57 @@ stdout; diagnostics remain on stderr.
 ## TUI commands
 
 ```text
-/providers   configure provider authentication and local endpoints
-/model       choose the main provider and model
-/workflow    choose the active workflow
-/agents      inspect agents and model overrides
-/project     switch the current project
-/resume      resume a saved session
-/memory      inspect and edit core memory
-/mcp         manage MCP servers and tools
-/skills      select a skill
-/channels    manage remote channels
-/compact     compact older context
-/setting     open settings
-/update      check for updates
-/doctor      diagnose installation health
+/clear        start a fresh chat
+/project      switch the current project
+/resume       resume a saved chat
+/compact      compact older conversation context
+/autoclear    manage idle-time context clearing
+/context      inspect the current context surface
+/usage        show provider quota and balance
+/providers    configure authentication and local endpoints
+/model        choose the main provider and model
+/websearch    choose the web search route
+/workflow     choose the active workflow
+/agents       inspect agents and model overrides
+/effort       set reasoning effort
+/fast         toggle supported model fast mode
+/OutputStyle  choose the Lead response style
+/theme        change the TUI color theme
+/memory       inspect and edit core memory
+/mcp          manage MCP servers and tools
+/skills       choose a skill for the next request
+/plugins      manage local plugin integrations
+/hooks        manage before-tool hooks and events
+/setting      open runtime settings
+/profile      set your title and response language
+/update       check for updates
+/doctor       diagnose installation health
+/quit         quit the TUI
 ```
 
 Workflows and agents are Markdown definition packs (`WORKFLOW.md`, `AGENT.md`).
 Built-in packs ship with Mixdog; custom packs live under the Mixdog data
 directory.
 
-## Windows desktop app
+## Desktop app
 
-Mixdog Desktop runs the same agent runtime as the CLI in an Electron
-workbench:
+Mixdog Desktop runs the same agent runtime as the CLI:
 
 - Split panes for parallel, independently routed agent sessions
+- Live session handoff between the TUI, desktop windows, and paired browsers
 - Monaco editor, LSP integration, diffs, and turn-by-turn edit review
 - Git staging, commits, branches, and generated commit messages
-- Windows file explorer with previews, thumbnails, and drag-and-drop
-- Integrated PowerShell and ConPTY terminal tabs
+- File explorer with previews, thumbnails, search, and drag-and-drop
+- Integrated terminal tabs using the local system shell
 - Image and video generation Studio with a persistent local gallery
-- Visual workflow, agent, and schedule editors
+- Visual workflow, agent, schedule, and webhook editors
+- Voice dictation with an optional local transcription runtime
 - Provider setup, usage, git identity, and remote pairing settings
 
+The paired remote web app is installable on desktop and mobile browsers. It
+uses an authenticated end-to-end encrypted connection before session state,
+terminal data, files, or operation requests cross the relay.
+
 For desktop development:
 
 ```bash
@@ -148,37 +218,6 @@ cd apps/desktop
 npm run dev
 ```
 
-## Terminal-Bench 2.1
-
-Head-to-head runs against other agent CLIs on the same 89 tasks, holding the
-model fixed and changing only the harness. Every run uses the official Harbor
-verifier, fast mode off, a 272k context window, and zero retries.
-
-**GPT-5.6 Sol xhigh — Mixdog vs Codex CLI**
-
-- Score: **86.5%** (385/445) vs Codex CLI's **84.3%** (75/89)
-- **1.11×** faster per trial (339s vs 378s)
-- **42%** lower priced cost ($0.641 vs $1.096 per trial)
-- **45%** smaller median final context (18.5k vs 33.5k tokens)
-
-**Claude Opus 5 — Mixdog vs Claude Code**
-
-- Score: **79/89** vs Claude Code's **77/89**
-- **1.15×** faster and **19%** lower priced cost ($104.29 vs $129.21 per run)
-- **28%** smaller median final context (27.6k vs 38.2k tokens)
-
-Measurement: the Mixdog Sol run follows the same protocol the official
-Terminal-Bench leaderboard requires — all 89 tasks repeated five times
-(`k=5`, 445 trials), scored by the official Harbor verifier. The Codex CLI
-baseline and both Opus-side runs are single passes (`k=1`, 89 trials each).
-
-The official leaderboard is not accepting community submissions, so every run
-here ships its raw artifacts instead — Harbor verdicts, official verifier
-output, pinned task checksums, and the usage snapshots behind every cost
-figure — alongside the harness, presets, and metric scripts that recompute
-each number above:
-[`benchmarks/terminal-bench-2.1/`](benchmarks/terminal-bench-2.1/).
-
 ## Data and configuration
 
 Mixdog uses `~/.mixdog` as its home root and `~/.mixdog/data` for runtime data
@@ -213,10 +252,11 @@ Main directories:
 
 ```text
 src/            CLI, TUI, runtime, workflows, agents, and rules
-apps/desktop/   Windows desktop workbench
-apps/relay/     remote web relay
+apps/desktop/   cross-platform desktop app
+apps/relay/     remote web app and relay
 native/         native process, search, patch, and support binaries
 scripts/        tests, diagnostics, benchmarks, and build scripts
+benchmarks/     reproducible benchmark harnesses, results, and raw artifacts
 vendor/         vendored runtime components
 ```
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "mixdog",
-  "version": "0.9.151",
+  "version": "0.9.153",
   "private": false,
   "type": "module",
   "description": "Standalone mixdog coding-agent CLI/TUI workspace.",
@@ -66,11 +66,17 @@
     "test:native-edit-wire": "node scripts/native-edit-wire-test.mjs",
     "test:release-critical": "npm run test:release-assets && npm run smoke:patch && npm run test:providers",
     "test:spec-advisory": "node scripts/advisory-spec-test.mjs && npm run test:spec-advisory --prefix apps/desktop",
-    "test:session-transport": "node --test scripts/session-transport-test.mjs scripts/daemon-bootstrap-test.mjs",
-    "test:session": "node --test scripts/runtime-turn-contract-test.mjs scripts/session-save-fault-store-test.mjs scripts/turn-review-revert-test.mjs src/runtime/shared/tool-surface.test.mjs",
+    "test:session-transport": "node --test scripts/session-transport-test.mjs scripts/daemon-bootstrap-test.mjs src/runtime/shared/child-spawn-remote.test.mjs src/standalone/session-runtime-agent-control-client.test.mjs src/standalone/session-runtime-dispatch-cancel.test.mjs src/standalone/session-runtime-host-factory.test.mjs src/standalone/session-runtime-inline-host.test.mjs src/standalone/session-runtime-provider-cooldown.test.mjs src/standalone/session-runtime-shard-host.test.mjs src/standalone/session-runtime-shard-router.test.mjs",
+    "test:session": "node --test scripts/runtime-turn-contract-test.mjs scripts/session-save-fault-store-test.mjs scripts/turn-review-revert-test.mjs src/runtime/shared/tool-surface.test.mjs src/runtime/bridge-clients.test.mjs",
+    "test:office": "node --test src/runtime/office/office-runtime.test.mjs src/runtime/office/office-assurance.test.mjs && npm run test:office-approval --prefix apps/desktop",
+    "test:office:live": "node scripts/run-office-live-tests.mjs",
     "test:media": "node --test src/runtime/media/store.test.mjs src/runtime/media/renditions.test.mjs src/runtime/media/adapters/codex-image.test.mjs",
     "failures": "node scripts/tool-failures.mjs",
     "trace:llm": "node scripts/llm-trace-summary.mjs",
+    "bench:office": "node src/runtime/office/benchmark.mjs",
+    "bench:office:contract": "node src/runtime/office/contract-benchmark.mjs",
+    "bench:office:assurance": "node src/runtime/office/assurance-benchmark.mjs",
+    "bench:office:quality:live": "node src/runtime/office/quality-live-benchmark.mjs",
     "diag:sessions": "node scripts/session-diag.mjs",
     "diag:tools": "node scripts/tool-efficiency-diag.mjs",
     "diag:sweep": "node scripts/session-sweep.mjs",
@@ -118,27 +124,34 @@
     "@huggingface/transformers": "^3.8.1",
     "@modelcontextprotocol/sdk": "^1.29.0",
     "@mozilla/readability": "^0.6.0",
+    "@napi-rs/canvas": "^1.0.8",
+    "@pdf-lib/fontkit": "^1.1.1",
     "chalk": "^5.6.2",
     "cli-highlight": "^2.1.11",
     "diff": "^9.0.0",
     "ink": "^7.1.0",
     "jsdom": "^27.0.0",
+    "jszip": "^3.10.1",
     "kiwi-nlp": "^0.21.0",
     "marked": "^14.1.4",
     "node-cron": "^4.5.0",
     "onnxruntime-node": "1.23.2",
     "openai": "^6.44.0",
+    "pdf-lib": "^1.17.1",
+    "pdfjs-dist": "^6.2.108",
     "pg": "^8.22.0",
     "puppeteer-core": "^25.2.0",
     "react": "^19.2.7",
     "sharp": "^0.35.3",
     "string-width": "^8.2.1",
     "strip-ansi": "^7.2.0",
+    "tesseract.js": "^7.0.0",
     "tiktoken": "^1.0.22",
-    "unpdf": "^1.8.0",
     "undici": "^8.5.0",
+    "unpdf": "^1.8.0",
     "wrap-ansi": "^10.0.0",
     "ws": "^8.21.0",
+    "yaml": "^2.9.0",
     "zod": "^3.25.76"
   },
   "overrides": {

package/scripts/computer-task-cost-report.mjs

@@ -0,0 +1,144 @@
+#!/usr/bin/env node
+import { mkdirSync, readFileSync, writeFileSync } from 'node:fs';
+import { dirname, resolve } from 'node:path';
+
+function arg(name, fallback) {
+  const prefix = `--${name}=`;
+  const inline = process.argv.find((value) => value.startsWith(prefix));
+  if (inline) return inline.slice(prefix.length);
+  const index = process.argv.indexOf(`--${name}`);
+  return index >= 0 && index + 1 < process.argv.length ? process.argv[index + 1] : fallback;
+}
+
+function load(path) {
+  return JSON.parse(readFileSync(resolve(path), 'utf8'));
+}
+
+function percentile(values, fraction) {
+  const sorted = values
+    .map(Number)
+    .filter(Number.isFinite)
+    .sort((a, b) => a - b);
+  if (sorted.length === 0) return 0;
+  return sorted[Math.max(0, Math.ceil(sorted.length * fraction) - 1)];
+}
+
+function fixed(value, digits = 2) {
+  return Number(Number(value || 0).toFixed(digits));
+}
+
+function distribution(values) {
+  return {
+    p50: percentile(values, 0.5),
+    p95: percentile(values, 0.95),
+  };
+}
+
+const modelPath = arg(
+  'model',
+  'artifacts/computer-use/computer-schema-sequence-guided-full.json',
+);
+const hostPath = arg(
+  'host',
+  'apps/desktop/artifacts/computer-use/scenario-repeat-further-optimized-v5.json',
+);
+const sequencePath = arg(
+  'sequence',
+  'artifacts/computer-use/sequence-performance-final.json',
+);
+const outputPath = resolve(arg(
+  'output',
+  'artifacts/computer-use/computer-task-cost-final.json',
+));
+
+const model = load(modelPath);
+const host = load(hostPath);
+const sequence = load(sequencePath);
+const rows = Array.isArray(model.rows) ? model.rows : [];
+const totalInputTokens = rows.map((row) => row.usage?.inputTokens);
+const mainInputTokens = rows.map((row) => row.usage?.mainInputTokens);
+const warmupInputTokens = rows.map((row) => row.usage?.warmupInputTokens);
+const cachedTokens = rows.map((row) => row.usage?.cachedTokens);
+const uncachedTokens = rows.map((row) => (
+  Number(row.usage?.inputTokens || 0) - Number(row.usage?.cachedTokens || 0)
+));
+const outputTokens = rows.map((row) => row.usage?.outputTokens);
+const modelLatency = rows.map((row) => row.duration_ms);
+const totalRuns = Number(host.total_runs || 0);
+const hostSummary = host.summary || {};
+const perToolCallInput = distribution(totalInputTokens);
+const perToolCallLatency = distribution(modelLatency);
+
+const taskProfile = (modelCalls) => ({
+  model_calls: modelCalls,
+  call_equivalent_input_tokens_p50: fixed(perToolCallInput.p50 * modelCalls),
+  call_equivalent_input_tokens_p95: fixed(perToolCallInput.p95 * modelCalls),
+  model_latency_p50_ms: fixed(perToolCallLatency.p50 * modelCalls),
+  note: 'Call-equivalent only; later turns can include fresh-state text or images and prompt-cache behavior.',
+});
+
+const report = {
+  schema_version: 1,
+  generated_at: new Date().toISOString(),
+  inputs: {
+    model: resolve(modelPath),
+    host: resolve(hostPath),
+    sequence: resolve(sequencePath),
+  },
+  model_call: {
+    samples: rows.length,
+    first_call_success_rate: model.summary?.first_call_success_rate ?? null,
+    total_input_tokens: perToolCallInput,
+    main_input_tokens: distribution(mainInputTokens),
+    warmup_input_tokens: distribution(warmupInputTokens),
+    cached_tokens: distribution(cachedTokens),
+    uncached_tokens: distribution(uncachedTokens),
+    output_tokens: distribution(outputTokens),
+    latency_ms: perToolCallLatency,
+  },
+  windows_host_scenario: {
+    samples: totalRuns,
+    pass_rate: hostSummary.success_rate ?? null,
+    model_facing_tool_calls_per_run: fixed(hostSummary.tool_calls / totalRuns, 3),
+    observations_per_run: fixed(hostSummary.observations / totalRuns, 3),
+    mutations_per_run: fixed(hostSummary.mutations / totalRuns, 3),
+    latency_ms: {
+      p50: hostSummary.scenario_duration_p50_ms,
+      p95: hostSummary.scenario_duration_p95_ms,
+    },
+    payload_bytes_per_run: {
+      request: fixed(hostSummary.request_bytes / totalRuns),
+      response_text: fixed(hostSummary.response_text_bytes / totalRuns),
+      images: fixed(hostSummary.image_bytes / totalRuns),
+    },
+  },
+  task_profiles: {
+    fresh_observation_action: taskProfile(1),
+    observe_then_action: taskProfile(2),
+    discover_observe_then_action: taskProfile(3),
+    safe_two_action_chain: {
+      separate: {
+        model_calls_after_observation: 2,
+        post_action_captures: sequence.separate?.post_action_captures,
+        host_latency_p50_ms: sequence.separate?.p50_ms,
+        host_latency_p95_ms: sequence.separate?.p95_ms,
+      },
+      sequence: {
+        model_calls_after_observation: 1,
+        post_action_captures: sequence.sequence?.post_action_captures,
+        host_latency_p50_ms: sequence.sequence?.p50_ms,
+        host_latency_p95_ms: sequence.sequence?.p95_ms,
+      },
+      reduction: sequence.reduction,
+    },
+  },
+  caveats: [
+    'Model token samples are provider-reported single-call usage from 36 representative first-call tasks.',
+    'Windows host samples are 230 real native, Electron, Chrome, Korean OCR, stale-state, and recovery runs.',
+    'Task profiles are call-equivalent projections, not billing claims; screenshots and growing conversation history can increase later-turn input.',
+  ],
+};
+
+mkdirSync(dirname(outputPath), { recursive: true });
+writeFileSync(outputPath, `${JSON.stringify(report, null, 2)}\n`, 'utf8');
+process.stdout.write(`${JSON.stringify(report, null, 2)}\n${outputPath}\n`);

package/scripts/run-office-live-tests.mjs

@@ -0,0 +1,32 @@
+import { spawn } from 'node:child_process';
+import { fileURLToPath } from 'node:url';
+
+const testPath = fileURLToPath(new URL('../src/runtime/office/office-live-runtime.test.mjs', import.meta.url));
+const child = spawn(process.execPath, ['--test', testPath], {
+  cwd: process.cwd(),
+  env: {
+    ...process.env,
+    MIXDOG_TEST_LIVE_OFFICE: '1',
+  },
+  stdio: ['inherit', 'pipe', 'pipe'],
+  windowsHide: true,
+});
+let tapFailed = false;
+let tail = '';
+const forward = (stream, target) => {
+  stream.on('data', (chunk) => {
+    const text = chunk.toString();
+    target.write(text);
+    tail = `${tail}${text}`.slice(-16_384);
+    if (/(?:^|\n)not ok \d+ -/m.test(tail)) tapFailed = true;
+  });
+};
+forward(child.stdout, process.stdout);
+forward(child.stderr, process.stderr);
+child.once('error', (error) => {
+  process.stderr.write(`${error?.stack || error}\n`);
+  process.exit(1);
+});
+child.once('close', (code) => {
+  process.exit(tapFailed ? 1 : Number.isInteger(code) ? code : 1);
+});

package/scripts/tool-stress.mjs

@@ -17,7 +17,6 @@ import { executePatchTool } from '../src/runtime/agent/orchestrator/tools/patch.
 import { normalizeToolEnvelope } from '../src/runtime/agent/orchestrator/session/tool-envelope.mjs';
 import { warmNativeSpawnServer } from '../src/runtime/agent/orchestrator/tools/lib/native-spawn-client.mjs';
 import { warmNativeSearchServer } from '../src/runtime/agent/orchestrator/tools/builtin/native-search-client.mjs';
-import { prewarmFindEnumeration } from '../src/runtime/agent/orchestrator/tools/builtin/list-tool.mjs';
 
 if (!process.argv.includes('--unsafe-live')) {
   console.error('Refusing high-impact tool stress without --unsafe-live; run `node scripts/tool-search-bench.mjs` for safe exploration diagnostics.');
@@ -66,7 +65,6 @@ try {
   await Promise.all([
     warmNativeSpawnServer(),
     warmNativeSearchServer(),
-    prewarmFindEnumeration(root),
   ]);
   prewarmMs = Date.now() - prewarmStarted;
   // ── Phase A+C: concurrent multi-session waves (search/read/graph/shell +

package/src/agents/front-worker/AGENT.md

@@ -0,0 +1,13 @@
+---
+name: Front Worker
+description: Use for web frontend, GUI, UI, visual presentation, and 2D or 3D asset implementation.
+---
+
+Own visual and interface implementation.
+
+Implement the requested visual outcome while preserving existing interaction
+behavior and project conventions. Keep changes targeted and maintain hierarchy,
+consistency, responsiveness, and accessibility when relevant. Preserve source
+formats and naming conventions for asset work.
+
+Hand off the completed visual outcome and changed files or assets.

package/src/agents/front-worker/agent.json

@@ -0,0 +1,6 @@
+{
+  "id": "front-worker",
+  "name": "Front Worker",
+  "description": "Use for web frontend, GUI, UI, visual presentation, and 2D or 3D asset implementation.",
+  "entry": "AGENT.md"
+}

package/src/agents/heavy-worker/AGENT.md

@@ -1,18 +1,17 @@
 ---
-permission: read-write
+name: Heavy Worker
+description: Use only for large refactors, cross-cutting architectural changes, difficult root-cause investigations, or implementation beyond the default Worker's normal scope.
 ---
 
-# Heavy Worker
-Own the assigned implementation slice through staged delivery.
+Own high-complexity implementation through staged delivery.
 
-Break work into bounded, dependency-aware slices and execute them in sequence.
-Keep the smallest coherent change; control blast radius rather than rewriting
-adjacent systems.
+Map the affected architecture and dependencies, divide the work into coherent
+stages, and execute them in dependency order. Preserve existing behavior unless
+the brief explicitly changes it, and control blast radius across boundaries.
 
-EDIT-FIRST DISCIPLINE. Patch incrementally and stop at the first explicit
-boundary: unclear ownership, a missing dependency, or growing blast radius.
-Do not cross that boundary without a new bounded assignment; report blocked
-work with the relevant file:line.
+When a required decision, dependency, or ownership boundary is unresolved,
+stop and report it with the relevant `file:line`.
 
-Finish the slice and report the changed `file:line`; verification belongs to the Lead and Reviewer.
+Hand off the completed outcome, material design decisions, and changed
+`file:line`.
 

package/src/agents/heavy-worker/agent.json

@@ -1,6 +1,6 @@
 {
   "id": "heavy-worker",
   "name": "Heavy Worker",
-  "description": "Use for complex, high-difficulty tasks.",
+  "description": "Use only for large refactors, cross-cutting architectural changes, difficult root-cause investigations, or implementation beyond the default Worker's normal scope.",
   "entry": "AGENT.md"
 }

package/src/agents/maintainer/AGENT.md

@@ -2,8 +2,6 @@
 permission: read-write
 ---
 
-# Maintainer
-
 Maintenance and cleanup agent.
 
 Smallest coherent change; no scope growth. Repeated read-only turns without

package/src/agents/reviewer/AGENT.md

@@ -1,19 +1,15 @@
 ---
-permission: read
+name: Reviewer
+description: Use after implementation to independently verify correctness, regressions, risks, and stated acceptance criteria before final reporting.
 ---
 
-# Reviewer
-
 Independent regression/risk review agent.
 
-Review the diff and tests with independent judgment. Prioritize actionable
-correctness, regression, security, and verification risks; inspect affected
-boundaries. Do not reimplement the change or report non-risky nits. Independently
-evaluate the final deliverable with a critical lens, actively seeking errors,
-unsupported assumptions, and counterexamples before confirming.
-Report findings first, severity-ordered, with one line per `file:line`. If clean,
-say so in one line and include only material residual risk.
+Inspect the diff, affected boundaries, existing tests, and stated acceptance
+criteria with independent judgment. Run the necessary builds, tests, lint, or
+runtime checks and actively seek regressions, unsupported assumptions, security
+risks, and counterexamples.
 
-When the work comes with stated criteria or reference material for judging
-it, verify against those as given — substituting your own interpretation or
-a self-built check is a verification risk to report.
+Do not modify files or reimplement the change. Report actionable findings first,
+severity-ordered, with evidence and one line per `file:line`. If clean, say so
+in one line and include only material residual risk.

package/src/agents/reviewer/agent.json

@@ -1,6 +1,6 @@
 {
   "id": "reviewer",
   "name": "Reviewer",
-  "description": "Use when an implementation finishes: cross-verify the change for correctness and risk before reporting.",
+  "description": "Use after implementation to independently verify correctness, regressions, risks, and stated acceptance criteria before final reporting.",
   "entry": "AGENT.md"
 }