create-walle 0.8.0 → 0.9.1

package/README.md

@@ -1,6 +1,12 @@
 # create-walle
 
-Set up **Wall-E** — your personal digital twin. An AI agent that learns from your Slack, email, and calendar to build a searchable second brain.
+Set up **CTM + Wall-E** in one command.
+
+**CTM** (Claude Task Manager) is a web-based dashboard for running and managing Claude Code sessions. It includes a terminal multiplexer, prompt editor, task queue, and code review panel — all in the browser.
+
+**Wall-E** is your personal digital twin. An AI agent that learns from your Slack, email, and calendar to build a searchable second brain. It comes with bundled skills (morning briefing, weekly reflection, proactive alerts, and more) and a chat interface with tool-use.
+
+Together they run as a single server: CTM on the main port, Wall-E on port+1.
 
 ## Install
 
@@ -13,11 +19,11 @@ This copies the project, installs dependencies, auto-detects your name and timez
 ## Commands
 
 ```bash
-npx create-walle install <dir>   # Install Wall-E (or update if already installed)
+npx create-walle install <dir>   # Install CTM + Wall-E (or update if already installed)
 npx create-walle update          # Update to latest version, keep your config
 npx create-walle start           # Start as background service (auto-restarts on reboot)
 npx create-walle stop            # Stop the service
-npx create-walle status          # Check if Wall-E is running
+npx create-walle status          # Check if running
 npx create-walle logs            # Tail the service logs
 npx create-walle uninstall       # Remove the auto-start service
 npx create-walle -v              # Show version
@@ -62,10 +68,20 @@ Wall-E API runs on `CTM_PORT + 1` (e.g., 5001).
 
 ## What's Included
 
-- **CTM Dashboard** — terminal multiplexer, prompt editor, task manager, code review
-- **Wall-E Agent** — ingest/think/reflect loops, 7 bundled skills, chat with tool-use
-- **Brain DB** — SQLite with full-text search across Slack, email, calendar
-- **Multi-device** — share data via Dropbox/iCloud, sessions tagged by hostname
+**CTM Dashboard:**
+- Terminal multiplexer — run multiple Claude Code sessions side by side
+- Prompt editor — save, organize, and queue prompts to Claude
+- Task queue — automated prompt execution with auto/manual modes
+- Code review panel — session insights and review workflows
+- Live session status — Running/Waiting/Idle indicators with scroll-up support during active output
+
+**Wall-E Agent:**
+- Ingest/think/reflect loops that continuously learn from your data
+- 12 bundled skills (morning briefing, weekly reflection, proactive alerts, slack mentions, email sync, and more)
+- Chat with tool-use — search memories, run skills, call MCP servers
+- Brain DB — SQLite with full-text search across Slack, email, calendar
+- Intelligence engine — context-aware compaction, initiative scoring, proactive briefings
+- Multi-device — share data via Dropbox/iCloud, sessions tagged by hostname
 
 ## Links
 

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "create-walle",
-  "version": "0.8.0",
-  "description": "Wall-E — your personal digital twin. AI agent that learns from Slack, email & calendar. Includes dashboard, chat, and 7 bundled skills.",
+  "version": "0.9.1",
+  "description": "CTM + Wall-E — Claude Task Manager dashboard and your personal digital twin AI agent. Terminal multiplexer, prompt editor, task queue, and an agent that learns from Slack, email & calendar.",
   "bin": {
     "create-walle": "bin/create-walle.js"
   },
@@ -15,5 +15,5 @@
     "prepublishOnly": "bash build.sh"
   },
   "license": "MIT",
-  "keywords": ["wall-e", "ai", "digital-twin", "claude", "personal-assistant"]
+  "keywords": ["ctm", "task-manager", "wall-e", "ai", "digital-twin", "claude", "personal-assistant"]
 }

package/template/CLAUDE.md

@@ -5,8 +5,21 @@
 NEVER kill CTM or Wall-E processes directly (e.g. `kill`, `lsof | xargs kill`, `pkill`).
 Always use the restart API or helper script — they handle graceful shutdown and auto-respawn.
 
-- **Restart CTM**: `bash claude-task-manager/bin/restart-ctm.sh` or `curl -sX POST http://localhost:3456/api/restart/ctm`
+- **Restart CTM**: `bash claude-task-manager/bin/restart-ctm.sh` or `curl -sX POST http://localhost:3456/api/restart/ctm?force=true`
 - **Restart Wall-E**: `curl -sX POST http://localhost:3456/api/restart/walle`
 - **Stop Wall-E**: `curl -sX POST http://localhost:3456/api/stop/walle`
 - **Start Wall-E**: `curl -sX POST http://localhost:3456/api/start/walle`
 - **Check status**: `curl -s http://localhost:3456/api/services/status`
+
+## CTM Development — CRITICAL SAFETY RULE
+
+**NEVER test code changes against the primary instance (port 3456/3457).**
+The primary instance is the user's live workbench — restarting or testing against it
+kills all active sessions. This applies to ALL operations: curl, fetch, restart, etc.
+
+When developing CTM or Wall-E code:
+1. Use the `/ctm-dev` skill (or run it manually)
+2. Pick a random port pair: `DEV_CTM_PORT=$((4100 + RANDOM % 900))`
+3. Start a dev instance: `DEV_CTM_PORT=$port DEV_WALLE_PORT=$((port+1)) bash bin/dev.sh`
+4. Test ONLY against your dev port, never 3456
+5. Only restart primary when the user explicitly asks — and ask for confirmation first

package/template/README.md

@@ -1,34 +1,54 @@
-# Wall-E
+# Wall-E + CTM
 
-**Your personal digital twin** — an AI agent that learns from your Slack, email, and calendar to build a searchable second brain, answer questions in your voice, and automate daily workflows.
+**Wall-E** is your personal digital twin — an AI agent that learns from your Slack, email, and calendar to build a searchable second brain, answer questions in your voice, and automate daily workflows.
 
-Wall-E runs locally alongside **CTM** (Claude Task Manager), a terminal multiplexer and task dashboard for Claude Code sessions.
+**CTM** (Claude Task Manager) is an agent task manager and terminal multiplexer — a web dashboard for running Claude Code sessions, managing prompts, reviewing code, and controlling Wall-E.
 
-## What It Does
+Together they give you a local-first AI command center: a browser-based workspace where you manage Claude Code sessions, chat with an AI that knows your work context, and automate the repetitive parts of your day.
 
-- **Ingests** your Slack messages, calendar events, and sent emails into a local SQLite brain
-- **Learns** your preferences, relationships, communication style, and knowledge over time
-- **Answers** questions about your work, meetings, and conversations — with citations
-- **Runs skills** on a schedule: morning briefings, Slack sync, calendar sync, and more
-- **Manages tasks** and Claude Code sessions through a web dashboard
+## What You Get
+
+### CTM — Agent Task Manager
+- **Terminal multiplexer** — run multiple Claude Code sessions side-by-side in browser tabs, with session history, search, and replay
+- **Prompt library** — rich-text editor with versioning, folders, tags, and one-click send to any active session
+- **Prompt queue** — batch multiple prompts and send them sequentially to a session with idle detection
+- **Shadow Approver** — learns your permission patterns and auto-approves safe operations across sessions
+- **Code review** — AI-assisted diff review across projects with inline comments
+- **Session insights** — AI analysis of your Claude Code usage with actionable recommendations
+- **Permission manager** — fine-grained control over what Claude Code can do automatically, with risk tiers and rule consolidation
+
+### Wall-E — Personal AI Agent
+- **Second brain** — ingests Slack messages, calendar events, and sent emails into a local SQLite database with full-text search
+- **Chat** — ask questions about your work, meetings, and conversations — grounded in your actual data with citations
+- **Scheduled skills** — morning briefings, Slack sync, calendar sync, email digest, and custom skills on configurable schedules
+- **Background tasks** — recurring or one-shot tasks with live logs, skill execution, and checkpoint/resume
+- **Knowledge graph** — automatically extracts facts, relationships, and patterns from your data over time
+- **People intelligence** — tracks relationships, communication patterns, trust levels, and personas
+- **MCP integration** — calls tools on Slack, GitHub, and custom MCP servers directly from chat
+- **macOS native** — calendar via EventKit, email via AppleScript, Spotlight file search, desktop notifications, clipboard access
 
 ## Architecture
 
 ```
-CTM (port 3456)                    Wall-E
+CTM (port 3456)                    Wall-E Daemon
 +-----------------------+          +------------------------+
 | Web Dashboard         |          | Agent Daemon           |
 |  - Sessions (pty)     |  <--->   |  - Ingest loop (60s)   |
 |  - Prompts editor     |          |  - Think loop (120s)   |
-|  - Task manager       |          |  - Reflect loop (1hr)  |
-|  - Code review        |          |  - Skills loop (5min)  |
-|  - Wall-E chat tab    |          |  - Tasks loop (30s)    |
-+-----------------------+          +------------------------+
-         |                                    |
-         v                                    v
-   task-manager.db                    wall-e-brain.db
-   (sessions, prompts,               (memories, knowledge,
-    tasks, reviews)                   people, skills, tasks)
+|  - Prompt queue       |          |  - Reflect loop (1hr)  |
+|  - Shadow Approver    |          |  - Skills loop (5min)  |
+|  - Code review        |          |  - Tasks loop (30s)    |
+|  - Session insights   |          |                        |
+|  - Permission manager |          | Chat Engine            |
+|  - Wall-E chat tab    |          |  - Tool use (local+MCP)|
++-----------------------+          |  - Memory search       |
+         |                         |  - Skill execution     |
+         v                         +------------------------+
+   task-manager.db                          |
+   (sessions, prompts,                      v
+    rules, reviews)              wall-e-brain.db
+                                 (memories, knowledge,
+                                  people, skills, tasks)
 ```
 
 ## Quickstart
@@ -39,7 +59,7 @@ cd my-agent
 node claude-task-manager/server.js
 ```
 
-Open **http://localhost:3456** — the browser setup page walks you through adding your API key and connecting integrations. Your name and timezone are auto-detected.
+Open **http://localhost:3456** — the setup page walks you through adding your API key and connecting integrations. Your name and timezone are auto-detected.
 
 Or clone manually:
 
@@ -52,6 +72,66 @@ node claude-task-manager/server.js
 
 The first run auto-creates `.env`, `wall-e-config.json`, and `~/.walle/data/`. Finish setup at http://localhost:3456/setup.html.
 
+## Dashboard Features
+
+### Sessions
+Terminal multiplexer for Claude Code — create, monitor, and manage multiple sessions from a single browser tab. Features include:
+- Live terminal output via WebSocket (xterm.js)
+- Session search and AI-powered search across all session history
+- AI auto-titling for sessions
+- Project grouping and filtering
+- Copy session command (with working directory)
+- Session replay and review
+
+### Prompts
+Rich-text prompt editor with a full formatting toolbar. Organize prompts into folders, tag them, track usage across sessions, and send to any active Claude Code session with one click.
+- **Composite prompts** — parent prompt with sub-prompts that compose into a single message
+- **Prompt queue** — batch multiple prompts, send sequentially with idle detection between items
+- **Power tools** — chains (multi-step sequences), templates, pattern detection, harvest (extract prompts from session history), and AI copilot suggestions
+- **Usage tracking** — see which sessions used which prompts and how often
+
+### Wall-E Chat
+Chat directly with Wall-E from the dashboard. Wall-E has access to your full brain (memories, knowledge, people) and can:
+- Search your Slack messages, emails, and calendar
+- Run skills on demand (morning briefing, Slack sync, etc.)
+- Call MCP tools (Slack, GitHub, etc.)
+- Create and manage background tasks
+- Execute shell commands, read files, take screenshots
+- Answer questions grounded in your actual data
+
+Sub-tabs: Chat, Tasks (background task manager), Skills (view and run), Brain (knowledge graph), Actions, Timeline, Questions, Status.
+
+### Shadow Approver
+Learns your permission patterns from Claude Code sessions and auto-approves safe operations. Features:
+- Pattern learning from your approval history
+- AI-driven approval decisions with confidence scoring
+- Domain-specific trust tiers (Observe, Draft, Guarded, Autonomous)
+- Rule consolidation (merges similar Bash rules)
+- Decision history and audit trail
+
+### Code Review
+AI-assisted code review across projects:
+- Inline diff viewer
+- Multi-project support
+- AI review comments with confidence levels
+- Send review to a Claude Code session for implementation
+
+### Session Insights
+AI analysis of your Claude Code usage patterns:
+- Workflow recommendations (automate repetitive tasks)
+- Prompt effectiveness analysis
+- Skill gap detection
+- Cost-saving suggestions
+- Session statistics
+
+### Permission Manager
+Fine-grained control over what Claude Code can do automatically:
+- Search and filter across all rules
+- Risk-based categorization (LOW, MEDIUM, HIGH)
+- Per-project or global scope
+- Rule consolidation for similar patterns
+- Import/export with Claude Code's settings files
+
 ## Configuration
 
 All configuration lives in `.env` (auto-generated on first run). Edit directly or use the browser setup page.
@@ -84,6 +164,7 @@ If you use `devbox ai -c claude`, Wall-E auto-reads your gateway credentials fro
 | `ANTHROPIC_AUTH_TOKEN` | No | Auth token when using a gateway |
 | `ANTHROPIC_CUSTOM_HEADERS_B64` | No | Base64-encoded custom headers (Portkey virtual keys, metadata) |
 | `WALLE_OWNER_NAME` | Auto | Your name (auto-detected from `git config`) |
+| `WALLE_MODEL` | Auto | Model selection (probed on setup: claude-sonnet-4-6, haiku-4-5, etc.) |
 | `CTM_PORT` | No | Dashboard port (default: `3456`) |
 | `WALL_E_PORT` | No | Wall-E API port (default: `CTM_PORT + 1`) |
 | `WALL_E_DATA_DIR` | No | Data directory (default: `~/.walle/data`) |
@@ -105,8 +186,9 @@ Wall-E ships with skills that run on a schedule to keep your brain up to date:
 | `slack-backfill` | manual | Full Slack history backfill (2022-present) |
 | `email-sync` | every 30m | Syncs sent emails from macOS Mail via JXA |
 | `email-digest` | daily 7am | Summarizes recent email activity |
-| `morning-briefing` | daily 7am | AI-generated daily briefing from all sources |
+| `morning-briefing` | daily 7am | AI-generated daily briefing: calendar, Slack activity, tasks, questions |
 | `memory-search` | on-demand | Full-text search across all memories |
+| `file-ingest` | manual | Read a folder of files into the brain with glob filtering |
 
 ### Creating Custom Skills
 
@@ -128,22 +210,15 @@ tags: [sync, data]
 
 Place skills in `wall-e/skills/_bundled/your-skill/SKILL.md` or load from a custom directory.
 
-## Dashboard Features
-
-### Sessions Tab
-Terminal multiplexer for Claude Code — create, monitor, and manage multiple sessions from a single browser tab.
+## Integrations
 
-### Prompts Tab
-Rich-text prompt editor with versioning, tagging, folders, and one-click send to any active Claude session. Supports composite prompts (parent + sub-prompts).
-
-### Tasks Tab
-Task dashboard with recurring task support, skill execution, checkpoint/resume, and live logs.
-
-### Wall-E Tab
-Chat directly with Wall-E — ask about your schedule, search your memories, run skills, and get AI-powered answers grounded in your actual data.
-
-### Code Review Tab
-Review code changes across projects with inline diffs and AI-assisted review.
+| Integration | How it works | Setup |
+|---|---|---|
+| **Slack** | OAuth + MCP protocol for search, read, send | Click "Connect" in setup page |
+| **Google Calendar** | macOS EventKit (reads all calendars: Google, iCloud, Outlook) | Automatic on macOS |
+| **Email** | macOS Mail via JXA/AppleScript | Automatic on macOS |
+| **GitHub** | MCP server (if configured in Claude Code) | Via MCP config |
+| **Custom MCP servers** | Any MCP-compatible server | Add to MCP config |
 
 ## API
 
@@ -152,17 +227,19 @@ CTM runs on port 3456. Key endpoints:
 | Method | Path | Description |
 |---|---|---|
 | GET | `/api/services/status` | CTM and Wall-E process status |
+| POST | `/api/restart/ctm` | Restart CTM server |
 | POST | `/api/restart/walle` | Restart Wall-E daemon |
 | POST | `/api/start/walle` | Start Wall-E daemon |
 | POST | `/api/stop/walle` | Stop Wall-E daemon |
-| GET | `/api/wall-e/status` | Wall-E brain stats and owner info |
+| GET | `/api/wall-e/status` | Brain stats, loop health, owner info |
 | GET | `/api/wall-e/memories` | List memories (filterable by source, since, limit) |
 | GET | `/api/wall-e/knowledge` | List knowledge entries |
 | GET | `/api/wall-e/people` | List known people |
 | GET | `/api/wall-e/timeline` | Memory timeline |
-| POST | `/api/wall-e/chat` | Chat with Wall-E |
+| GET | `/api/wall-e/tasks` | List background tasks |
+| POST | `/api/wall-e/chat` | Chat with Wall-E (supports SSE streaming) |
 | GET | `/api/prompts` | List prompts |
-| GET | `/api/sessions` | List active terminal sessions |
+| GET | `/api/recent-sessions` | List recent Claude Code sessions |
 
 ## Tech Stack
 
@@ -170,24 +247,27 @@ CTM runs on port 3456. Key endpoints:
 - **Database**: SQLite via better-sqlite3 (WAL mode, FTS5 full-text search)
 - **AI**: Claude API via Anthropic SDK (supports Portkey gateway)
 - **Frontend**: Vanilla HTML/CSS/JS (no build step, no React)
+- **Terminal**: xterm.js (frontend) + node-pty (backend)
 - **macOS integration**: EventKit (calendar), JXA (email/AppleScript), Spotlight (file search)
-- **Terminal**: node-pty for session management
 
 ## Development
 
 ```bash
+# Start dev instance (doesn't affect primary on :3456)
+bash bin/dev.sh
+
 # Run Wall-E tests
 cd wall-e && npm test
 
-# Run a single test file
-node --test tests/brain.test.js
-
 # Run Slack backfill manually
 node scripts/slack-backfill.js 2024-01  # single month
 node scripts/slack-backfill.js incremental  # new messages only
 
 # Run calendar sync manually
 node skills/_bundled/google-calendar/run.js
+
+# Run morning briefing manually
+node skills/_bundled/morning-briefing/run.js
 ```
 
 ## License

package/template/claude-task-manager/bin/restart-ctm.sh

@@ -1,12 +1,13 @@
 #!/bin/bash
 # Graceful CTM restart — spawns a watcher process, then exits cleanly.
 # Use this instead of killing the process directly.
-curl -s -X POST "http://localhost:3456/api/restart/ctm" | cat
+CTM_PORT="${CTM_PORT:-3456}"
+curl -s -X POST "http://localhost:${CTM_PORT}/api/restart/ctm?force=true" | cat
 echo ""
 echo "CTM server restarting... waiting for it to come back."
 sleep 2
 for i in $(seq 1 15); do
-  if curl -sf "http://localhost:3456/api/services/status" > /dev/null 2>&1; then
+  if curl -sf "http://localhost:${CTM_PORT}/api/services/status" > /dev/null 2>&1; then
     echo "CTM server is back up."
     exit 0
   fi

package/template/claude-task-manager/db.js

@@ -594,6 +594,21 @@ function runMigrations() {
   if (!scCols2.find(c => c.name === 'hostname')) {
     getDb().prepare("ALTER TABLE session_conversations ADD COLUMN hostname TEXT DEFAULT ''").run();
   }
+
+  // startup_tasks: tracks active sessions for crash-safe restore on restart
+  const hasStartupTasks = getDb().prepare("SELECT name FROM sqlite_master WHERE type='table' AND name='startup_tasks'").get();
+  if (!hasStartupTasks) {
+    getDb().exec(`
+      CREATE TABLE startup_tasks (
+        session_id TEXT PRIMARY KEY,
+        label      TEXT NOT NULL DEFAULT '',
+        cmd        TEXT NOT NULL DEFAULT '',
+        args       TEXT NOT NULL DEFAULT '[]',
+        cwd        TEXT NOT NULL DEFAULT '',
+        created_at TEXT NOT NULL DEFAULT (datetime('now'))
+      )
+    `);
+  }
 }
 
 // --- Settings CRUD ---
@@ -1115,9 +1130,9 @@ function listSessionConversations({ search, limit, offset, hostname, allDevices
     params.push(hostname);
   }
   if (search) {
-    sql += ' AND (title LIKE ? OR first_message LIKE ? OR project_path LIKE ?)';
+    sql += ' AND (title LIKE ? OR first_message LIKE ? OR project_path LIKE ? OR messages LIKE ?)';
     const q = `%${search}%`;
-    params.push(q, q, q);
+    params.push(q, q, q, q);
   }
   sql += ' ORDER BY imported_at DESC';
   if (limit) { sql += ' LIMIT ?'; params.push(limit); }
@@ -1699,6 +1714,28 @@ function getInsightsData(includeInternal) {
   };
 }
 
+// --- Startup Tasks (crash-safe session restore) ---
+function addStartupTask(sessionId, label, cmd, args, cwd) {
+  getDb().prepare(
+    'INSERT OR REPLACE INTO startup_tasks (session_id, label, cmd, args, cwd) VALUES (?, ?, ?, ?, ?)'
+  ).run(sessionId, label || '', cmd || '', JSON.stringify(args || []), cwd || '');
+}
+
+function removeStartupTask(sessionId) {
+  getDb().prepare('DELETE FROM startup_tasks WHERE session_id = ?').run(sessionId);
+}
+
+function listStartupTasks() {
+  return getDb().prepare('SELECT * FROM startup_tasks ORDER BY created_at').all().map(row => ({
+    ...row,
+    args: JSON.parse(row.args || '[]'),
+  }));
+}
+
+function clearStartupTasks() {
+  getDb().prepare('DELETE FROM startup_tasks').run();
+}
+
 module.exports = {
   initDb, getDb, closeDb, getDbPath: () => currentDbPath,
   getSetting, getSettingsByPrefix, setSetting,
@@ -1729,5 +1766,6 @@ module.exports = {
   replaceInsightRecommendations, listInsightRecommendations,
   startAnalysisRun, completeAnalysisRun, getLastAnalysisRun,
   getInsightsData,
+  addStartupTask, removeStartupTask, listStartupTasks, clearStartupTasks,
   DEFAULT_IMAGES_DIR, BACKUP_DIR,
 };

package/template/claude-task-manager/public/css/walle.css

@@ -7,6 +7,7 @@
   border-bottom: 1px solid var(--border); flex-shrink: 0;
 }
 .walle-header-title { font-size: 15px; font-weight: 700; color: var(--fg); }
+.walle-avatar { vertical-align: middle; margin-right: 2px; position: relative; top: -1px; }
 
 .walle-subnav { display: flex; gap: 4px; margin-left: auto; }
 .walle-subnav-btn {
@@ -541,6 +542,25 @@
 .we-task-group[open] > .we-task-group-header::before { transform: rotate(90deg); }
 .we-task-group-header:hover { background: rgba(255,255,255,0.03); }
 
+/* View tabs */
+.we-view-tabs {
+  display: flex; align-items: center; gap: 2px;
+  padding: 4px 10px 0; border-bottom: 1px solid var(--border, #2a2a3e);
+}
+.we-view-tab {
+  padding: 6px 14px; border: none; background: transparent;
+  color: var(--fg-muted, #888); font-size: 12px; font-weight: 500;
+  cursor: pointer; border-bottom: 2px solid transparent;
+  border-radius: 4px 4px 0 0; transition: all 0.15s;
+}
+.we-view-tab:hover { color: var(--fg, #e0e0e0); background: rgba(255,255,255,0.04); }
+.we-view-tab.active { color: var(--accent, #60a5fa); border-bottom-color: var(--accent, #60a5fa); font-weight: 600; }
+.we-view-tab-badge {
+  display: inline-block; min-width: 16px; height: 16px; line-height: 16px;
+  text-align: center; font-size: 10px; font-weight: 700; border-radius: 8px;
+  background: #e03131; color: #fff; margin-left: 4px; padding: 0 4px;
+}
+
 /* Task card */
 .we-task-card {
   background: var(--bg-lighter, #1e1e2e);
@@ -549,7 +569,10 @@
   border-radius: 6px;
   padding: 10px 14px;
   margin: 6px 10px;
+  transition: padding 0.15s;
 }
+.we-task-card.compact { padding: 6px 14px; }
+.we-task-card.compact .we-task-card-header { margin-bottom: 0; }
 .we-task-card-header {
   display: flex; justify-content: space-between; align-items: center;
   margin-bottom: 4px;
@@ -575,6 +598,106 @@
 .we-task-meta-extra { display: none; }
 .we-task-meta.we-meta-expanded .we-task-meta-extra { display: contents; }
 
+/* Compact metadata line */
+.we-task-meta-compact {
+  display: flex; align-items: center; gap: 10px; flex-wrap: wrap;
+  font-size: 11px; color: #777; margin-top: 2px;
+}
+.we-task-meta-compact span { white-space: nowrap; }
+
+/* Source icons */
+.we-src-icon {
+  display: inline-block; padding: 1px 5px; border-radius: 3px;
+  font-size: 9px; font-weight: 600; text-transform: uppercase; letter-spacing: 0.3px;
+}
+.we-src-icon.slack { background: #1a2e1a; color: #4ade80; }
+.we-src-icon.chat { background: #1a1a3e; color: #60a5fa; }
+.we-src-icon.init { background: #2e1a2e; color: #c084fc; }
+.we-src-link {
+  color: #60a5fa; text-decoration: none; font-size: 11px;
+}
+.we-src-link:hover { text-decoration: underline; }
+
+/* Slack thread info panel */
+.we-slack-info { margin: 8px 0 4px; }
+.we-slack-thread-status {
+  display: flex; align-items: center; gap: 8px;
+  padding: 6px 10px; margin-top: 6px;
+  background: rgba(255,255,255,0.03); border-radius: 6px;
+  font-size: 12px;
+}
+.we-thread-badge {
+  font-weight: 600; font-size: 11px;
+  padding: 2px 8px; border-radius: 10px;
+}
+.we-thread-badge.active { background: #1a2e1a; color: #4ade80; }
+.we-thread-badge.expired { background: #2a2a2a; color: #888; }
+.we-thread-expires { color: #888; font-size: 11px; }
+
+/* Compact inline action buttons */
+.we-compact-actions { display: inline-flex; gap: 4px; margin-left: auto; }
+.we-compact-btn {
+  width: 22px; height: 22px; border-radius: 4px; border: 1px solid rgba(255,255,255,0.1);
+  background: transparent; color: #888; font-size: 11px; cursor: pointer;
+  display: inline-flex; align-items: center; justify-content: center;
+  transition: all 0.15s;
+}
+.we-compact-btn:hover { background: rgba(255,255,255,0.08); color: #ccc; }
+.we-compact-btn.run:hover { color: #4ade80; border-color: #4ade80; }
+.we-compact-btn.stop:hover { color: #e03131; border-color: #e03131; }
+
+/* Task context menu */
+.we-task-menu {
+  background: #1e1e2e; border: 1px solid rgba(255,255,255,0.15); border-radius: 6px;
+  min-width: 140px; padding: 4px 0; z-index: 1000;
+  box-shadow: 0 4px 16px rgba(0,0,0,0.5);
+}
+.we-task-menu-item {
+  padding: 6px 12px; font-size: 12px; color: #ccc; cursor: pointer;
+  white-space: nowrap;
+}
+.we-task-menu-item:hover { background: rgba(255,255,255,0.08); color: #fff; }
+.we-task-menu-item.danger { color: #e03131; }
+.we-task-menu-item.danger:hover { background: rgba(224,49,49,0.15); }
+
+/* Automation table view */
+.we-auto-health {
+  display: flex; gap: 12px; padding: 8px 10px; margin-bottom: 4px;
+  font-size: 12px; font-weight: 500;
+}
+.we-auto-health-item.ok { color: #5c940d; }
+.we-auto-health-item.paused { color: #888; }
+.we-auto-health-item.failed { color: #e03131; }
+
+.we-auto-table {
+  width: 100%; border-collapse: collapse; font-size: 12px;
+}
+.we-auto-table th {
+  text-align: left; padding: 6px 10px; color: #666; font-weight: 500;
+  font-size: 10px; text-transform: uppercase; letter-spacing: 0.5px;
+  border-bottom: 1px solid var(--border, #2a2a3e);
+}
+.we-auto-row {
+  cursor: pointer; transition: background 0.1s;
+}
+.we-auto-row:hover { background: rgba(255,255,255,0.03); }
+.we-auto-row.expanded { background: rgba(59,130,246,0.05); }
+.we-auto-row td { padding: 8px 10px; border-bottom: 1px solid rgba(255,255,255,0.03); }
+.we-auto-title { color: var(--fg, #e0e0e0); font-weight: 500; }
+.we-auto-sched { color: #888; }
+.we-auto-last { color: #777; }
+.we-auto-actions { text-align: right; }
+.we-auto-btn {
+  padding: 2px 8px; font-size: 10px; border-radius: 3px; cursor: pointer;
+  border: 1px solid rgba(255,255,255,0.1); background: transparent; color: #888;
+}
+.we-auto-btn:hover { background: rgba(255,255,255,0.08); color: #ccc; }
+.we-auto-detail td {
+  padding: 0 10px 12px 10px;
+  background: rgba(59,130,246,0.03);
+  border-bottom: 1px solid var(--border, #2a2a3e);
+}
+
 /* Chat date separator */
 .we-chat-date-sep { text-align: center; color: var(--fg-dim, #666); font-size: 11px; padding: 12px 0 4px; position: relative; }
 .we-chat-date-sep::before { content: ''; position: absolute; left: 0; right: 0; top: 50%; border-top: 1px solid rgba(255,255,255,0.06); }