oh-my-claude-sisyphus 3.5.7 → 3.6.0

package/commands/swarm.md

@@ -1,5 +1,5 @@
 ---
-description: N coordinated agents with atomic task claiming from shared pool
+description: N coordinated agents on shared task list with SQLite-based atomic claiming
 aliases: [swarm-agents]
 ---
 
@@ -7,7 +7,7 @@ aliases: [swarm-agents]
 
 [SWARM MODE ACTIVATED]
 
-Spawn N coordinated agents working on a shared task list with atomic claiming. Like a dev team tackling multiple files in parallel.
+Spawn N coordinated agents working on a shared task list with SQLite-based atomic claiming. Like a dev team tackling multiple files in parallel—fast, reliable, and with full fault tolerance.
 
 ## User's Request
 
@@ -22,22 +22,22 @@ Spawn N coordinated agents working on a shared task list with atomic claiming. L
 ### Parameters
 
 - **N** - Number of agents (1-5, enforced by Claude Code limit)
-- **agent-type** - Agent to spawn (executor, build-fixer, designer, etc.)
+- **agent-type** - Agent to spawn (e.g., executor, build-fixer, architect)
 - **task** - High-level task to decompose and distribute
 
 ### Examples
 
-```
+```bash
 /oh-my-claudecode:swarm 5:executor "fix all TypeScript errors"
 /oh-my-claudecode:swarm 3:build-fixer "fix build errors in src/"
 /oh-my-claudecode:swarm 4:designer "implement responsive layouts for all components"
 /oh-my-claudecode:swarm 2:architect "analyze and document all API endpoints"
 ```
 
-## How It Works
+## Architecture
 
 ```
-User: /swarm 5:executor "fix all TypeScript errors"
+User: "/swarm 5:executor fix all TypeScript errors"
               |
               v
       [SWARM ORCHESTRATOR]
@@ -50,14 +50,29 @@ User: /swarm 5:executor "fix all TypeScript errors"
    +--+--+--+--+
           |
           v
-   [SHARED TASK LIST]
-   - Fix a.ts (claimed E1)
-   - Fix b.ts (done E2)
-   - Fix c.ts (claimed E3)
-   - Fix d.ts (pending)
-   ...
+    [SQLITE DATABASE]
+    ┌─────────────────────┐
+    │ tasks table         │
+    ├─────────────────────┤
+    │ id, description     │
+    │ status (pending,    │
+    │   claimed, done,    │
+    │   failed)           │
+    │ claimed_by, claimed_at
+    │ completed_at, result│
+    │ error               │
+    ├─────────────────────┤
+    │ heartbeats table    │
+    │ (agent monitoring)  │
+    └─────────────────────┘
 ```
 
+**Key Features:**
+- SQLite transactions ensure only one agent can claim a task
+- Lease-based ownership with automatic timeout and recovery
+- Heartbeat monitoring for detecting dead agents
+- Full ACID compliance for task state
+
 ## Workflow
 
 ### 1. Parse Input
@@ -67,161 +82,329 @@ From `{{ARGUMENTS}}`, extract:
 - agent-type (executor, build-fixer, etc.)
 - task description
 
-### 2. Create Task List
-
-1. Analyze codebase based on task
-2. Break into file-specific subtasks
-3. Initialize `.omc/state/swarm-tasks.json` with all subtasks
-4. Each task gets: id, file, description, status, owner, timestamp
+### 2. Create Task Pool
+- Analyze codebase based on task
+- Break into file-specific subtasks
+- Initialize SQLite database with task pool
+- Each task gets: id, description, status (pending), and metadata columns
 
 ### 3. Spawn Agents
-
-Launch N agents via Task tool:
+- Launch N agents via Task tool
 - Set `run_in_background: true` for all
-- Each agent receives:
-  - Reference to shared task list
-  - Claiming protocol instructions
-  - Completion criteria
+- Each agent connects to the SQLite database
+- Agents enter claiming loop automatically
+
+**Important:** Use worker preamble when spawning agents to prevent sub-agent recursion:
 
-### 4. Task Claiming Protocol
+```typescript
+import { wrapWithPreamble } from '../agents/preamble.js';
 
+const prompt = wrapWithPreamble(`Your task: ${taskDescription}`);
+```
+
+### 4. Task Claiming Protocol (SQLite Transactional)
 Each agent follows this loop:
 
 ```
 LOOP:
-  1. Read swarm-tasks.json
-  2. Find first task with status="pending"
-  3. Atomically claim task (set status="claimed", add owner, timestamp)
-  4. Execute task
-  5. Mark task as "done"
-  6. GOTO LOOP (until no pending tasks)
+  1. Call claimTask(agentId)
+  2. SQLite transaction:
+     - Find first pending task
+     - UPDATE status='claimed', claimed_by=agentId, claimed_at=now
+     - INSERT/UPDATE heartbeat record
+     - Atomically commit (only one agent succeeds)
+  3. Execute task
+  4. Call completeTask(agentId, taskId, result) or failTask()
+  5. GOTO LOOP (until hasPendingWork() returns false)
+```
+
+**Atomic Claiming Details:**
+- SQLite `IMMEDIATE` transaction prevents race conditions
+- Only agent updating the row successfully gets the task
+- Heartbeat automatically updated on claim
+- If claim fails (already claimed), agent retries with next task
+- Lease Timeout: 5 minutes per task
+- If timeout exceeded + no heartbeat, cleanupStaleClaims releases task back to pending
+
+### 5. Heartbeat Protocol
+- Agents call `heartbeat(agentId)` every 60 seconds (or custom interval)
+- Heartbeat records: agent_id, last_heartbeat timestamp, current_task_id
+- Orchestrator runs cleanupStaleClaims every 60 seconds
+- If heartbeat is stale (>5 minutes old) and task claimed, task auto-releases
+
+### 6. Progress Tracking
+- Orchestrator monitors via TaskOutput
+- Shows live progress: pending/claimed/done/failed counts
+- Active agent count via getActiveAgents()
+- Reports which agent is working on which task via getAgentTasks()
+- Detects idle agents (all tasks claimed by others)
+
+### 7. Completion
+Exit when ANY of:
+- isSwarmComplete() returns true (all tasks done or failed)
+- All agents idle (no pending tasks, no claimed tasks)
+- User cancels via `/oh-my-claudecode:cancel`
+
+## Storage
+
+### SQLite Database (`.omc/state/swarm.db`)
+
+The swarm uses a single SQLite database stored at `.omc/state/swarm.db`. This provides:
+- **ACID compliance** - All task state transitions are atomic
+- **Concurrent access** - Multiple agents query/update safely
+- **Persistence** - State survives agent crashes
+- **Query efficiency** - Fast status lookups and filtering
+
+#### `tasks` Table Schema
+```sql
+CREATE TABLE tasks (
+  id TEXT PRIMARY KEY,
+  description TEXT NOT NULL,
+  status TEXT NOT NULL DEFAULT 'pending',
+    -- pending: waiting to be claimed
+    -- claimed: claimed by an agent, in progress
+    -- done: completed successfully
+    -- failed: completed with error
+  claimed_by TEXT,                  -- agent ID that claimed this task
+  claimed_at INTEGER,               -- Unix timestamp when claimed
+  completed_at INTEGER,             -- Unix timestamp when completed
+  result TEXT,                      -- Optional result/output from task
+  error TEXT                        -- Error message if task failed
+);
+```
+
+#### `heartbeats` Table Schema
+```sql
+CREATE TABLE heartbeats (
+  agent_id TEXT PRIMARY KEY,
+  last_heartbeat INTEGER NOT NULL,  -- Unix timestamp of last heartbeat
+  current_task_id TEXT              -- Task agent is currently working on
+);
 ```
 
-**Atomic Claiming:**
-- Read current task status
-- If "pending", claim it (set owner, timeout)
-- If already claimed, try next task
-- Timeout: 5 minutes per task
-- Timed-out tasks auto-release to "pending"
-
-### 5. Progress Tracking
-
-Orchestrator monitors via TaskOutput:
-- Shows live stats: claimed/done/pending counts
-- Reports which agent is working on which file
-- Detects idle agents (no pending tasks)
-
-### 6. Completion
-
-Exit when:
-- All tasks marked "done"
-- All agents idle (no pending tasks)
-- User cancels via `/cancel`
-
-## State Files
-
-### `.omc/swarm-state.json`
-Session-level state:
-
-```json
-{
-  "session_id": "swarm-20260124-143022",
-  "agent_count": 5,
-  "agent_type": "executor",
-  "task_description": "fix all TypeScript errors",
-  "status": "active",
-  "started_at": "2026-01-24T14:30:22Z",
-  "agents": [
-    {"id": "agent-1", "background_task_id": "task_abc", "status": "working"},
-    {"id": "agent-2", "background_task_id": "task_def", "status": "working"}
-  ]
+#### `swarm_session` Table Schema
+```sql
+CREATE TABLE swarm_session (
+  id INTEGER PRIMARY KEY CHECK (id = 1),
+  session_id TEXT NOT NULL,
+  active INTEGER NOT NULL DEFAULT 1,
+  agent_count INTEGER NOT NULL,
+  started_at INTEGER NOT NULL,
+  completed_at INTEGER
+);
+```
+
+## Task Claiming Protocol (Detailed)
+
+### Atomic Claim Operation with SQLite
+
+The core strength of the implementation is transactional atomicity:
+
+```typescript
+function claimTask(agentId: string): ClaimResult {
+  // Transaction ensures only ONE agent succeeds
+  const claimTransaction = db.transaction(() => {
+    // Step 1: Find first pending task
+    const task = db.prepare(
+      'SELECT id, description FROM tasks WHERE status = "pending" ORDER BY id LIMIT 1'
+    ).get();
+
+    if (!task) {
+      return { success: false, reason: 'No pending tasks' };
+    }
+
+    // Step 2: Attempt claim (will only succeed if status is still 'pending')
+    const result = db.prepare(
+      'UPDATE tasks SET status = "claimed", claimed_by = ?, claimed_at = ? WHERE id = ? AND status = "pending"'
+    ).run(agentId, Date.now(), task.id);
+
+    if (result.changes === 0) {
+      // Another agent claimed it between SELECT and UPDATE - try next
+      return { success: false, reason: 'Task was claimed by another agent' };
+    }
+
+    // Step 3: Update heartbeat to show we're alive and working
+    db.prepare(
+      'INSERT OR REPLACE INTO heartbeats (agent_id, last_heartbeat, current_task_id) VALUES (?, ?, ?)'
+    ).run(agentId, Date.now(), task.id);
+
+    return { success: true, taskId: task.id, description: task.description };
+  });
+
+  return claimTransaction();  // Atomic execution
 }
 ```
 
-### `.omc/state/swarm-tasks.json`
-Shared task list:
-
-```json
-{
-  "tasks": [
-    {
-      "id": "task-001",
-      "file": "src/utils/validation.ts",
-      "description": "Fix type errors in validation helpers",
-      "status": "claimed",
-      "owner": "agent-1",
-      "claimed_at": "2026-01-24T14:30:25Z",
-      "timeout_at": "2026-01-24T14:35:25Z"
-    },
-    {
-      "id": "task-002",
-      "file": "src/components/Header.tsx",
-      "description": "Fix missing prop types",
-      "status": "done",
-      "owner": "agent-2",
-      "claimed_at": "2026-01-24T14:30:26Z",
-      "completed_at": "2026-01-24T14:32:15Z"
-    },
-    {
-      "id": "task-003",
-      "file": "src/api/client.ts",
-      "description": "Add return type annotations",
-      "status": "pending",
-      "owner": null
+**Why SQLite Transactions Work:**
+- `db.transaction()` uses `IMMEDIATE` locking
+- Prevents other agents from modifying rows between SELECT and UPDATE
+- All-or-nothing atomicity: claim succeeds completely or fails completely
+- No race conditions, no lost updates
+
+### Lease Timeout & Auto-Release
+
+Tasks are automatically released if claimed too long without heartbeat:
+
+```typescript
+function cleanupStaleClaims(leaseTimeout: number = 5 * 60 * 1000) {
+  // Default 5-minute timeout
+  const cutoffTime = Date.now() - leaseTimeout;
+
+  const cleanupTransaction = db.transaction(() => {
+    // Find claimed tasks where:
+    // 1. Claimed longer than timeout, OR
+    // 2. Agent hasn't sent heartbeat in that time
+    const staleTasks = db.prepare(`
+      SELECT t.id
+      FROM tasks t
+      LEFT JOIN heartbeats h ON t.claimed_by = h.agent_id
+      WHERE t.status = 'claimed'
+        AND t.claimed_at < ?
+        AND (h.last_heartbeat IS NULL OR h.last_heartbeat < ?)
+    `).all(cutoffTime, cutoffTime);
+
+    // Release each stale task back to pending
+    for (const staleTask of staleTasks) {
+      db.prepare('UPDATE tasks SET status = "pending", claimed_by = NULL, claimed_at = NULL WHERE id = ?')
+        .run(staleTask.id);
     }
-  ],
-  "stats": {
-    "total": 15,
-    "pending": 8,
-    "claimed": 5,
-    "done": 2
-  }
+
+    return staleTasks.length;
+  });
+
+  return cleanupTransaction();
 }
 ```
 
-## Agent Instructions Template
+**How Recovery Works:**
+1. Orchestrator calls cleanupStaleClaims() every 60 seconds
+2. If agent hasn't sent heartbeat in 5 minutes, task is auto-released
+3. Another agent picks up the orphaned task
+4. Original agent can continue working (it doesn't know it was released)
+5. When original agent tries to mark task as done, verification fails safely
+
+## API Reference
+
+### Core API Functions
+
+#### `startSwarm(config: SwarmConfig): Promise<boolean>`
+Initialize the swarm with task pool and start cleanup timer.
+
+#### `stopSwarm(deleteDatabase?: boolean): boolean`
+Stop the swarm and optionally delete the database.
+
+#### `claimTask(agentId: string): ClaimResult`
+Claim the next pending task. Returns `{ success, taskId, description, reason }`.
+
+#### `completeTask(agentId: string, taskId: string, result?: string): boolean`
+Mark a task as done. Only succeeds if agent still owns the task.
 
-Each spawned agent receives:
+#### `failTask(agentId: string, taskId: string, error: string): boolean`
+Mark a task as failed with error details.
 
-```markdown
-You are agent {id} in a swarm of {N} {agent-type} agents.
+#### `heartbeat(agentId: string): boolean`
+Send a heartbeat to indicate agent is alive. Call every 60 seconds during long-running tasks.
 
-**Your Task:** {task_description}
+#### `cleanupStaleClaims(leaseTimeout?: number): number`
+Manually trigger cleanup of expired claims. Called automatically every 60 seconds.
 
-**Shared Task List:** .omc/state/swarm-tasks.json
+#### `hasPendingWork(): boolean`
+Check if there are unclaimed tasks available.
 
-**Your Loop:**
-1. Read swarm-tasks.json
-2. Find first task with status="pending"
-3. Claim it atomically (set status="claimed", owner="{id}", timeout)
-4. Execute the task
-5. Mark status="done", set completed_at
-6. Repeat until no pending tasks
+#### `isSwarmComplete(): boolean`
+Check if all tasks are done or failed.
 
-**Claiming Protocol:**
-- Read file, check status="pending"
-- Update status="claimed", add your ID
-- Set timeout_at = now + 5 minutes
-- Write file back
-- If file changed between read/write, retry
+#### `getSwarmStats(): SwarmStats | null`
+Get task counts and timing info.
 
-**Completion:**
-When no pending tasks remain, exit cleanly.
+### Configuration (SwarmConfig)
+
+```typescript
+interface SwarmConfig {
+  agentCount: number;           // Number of agents (1-5)
+  tasks: string[];              // Task descriptions
+  agentType?: string;           // Agent type (default: 'executor')
+  leaseTimeout?: number;        // Milliseconds (default: 5 min)
+  heartbeatInterval?: number;   // Milliseconds (default: 60 sec)
+  cwd?: string;                 // Working directory
+}
 ```
 
-## Constraints
+### Types
+
+```typescript
+interface SwarmTask {
+  id: string;
+  description: string;
+  status: 'pending' | 'claimed' | 'done' | 'failed';
+  claimedBy: string | null;
+  claimedAt: number | null;
+  completedAt: number | null;
+  error?: string;
+  result?: string;
+}
 
-- **Max Agents:** 5 (Claude Code background task limit)
-- **Claim Timeout:** 5 minutes per task
-- **Auto-Release:** Timed-out claims automatically released
-- **Heartbeat:** Recommended every 60 seconds
+interface ClaimResult {
+  success: boolean;
+  taskId: string | null;
+  description?: string;
+  reason?: string;
+}
 
-## Error Handling
+interface SwarmStats {
+  totalTasks: number;
+  pendingTasks: number;
+  claimedTasks: number;
+  doneTasks: number;
+  failedTasks: number;
+  activeAgents: number;
+  elapsedTime: number;
+}
+```
 
-- **Agent Crash:** Task auto-releases after timeout
-- **State Corruption:** Orchestrator validates and repairs
-- **No Pending Tasks:** Agent exits cleanly
-- **All Agents Idle:** Orchestrator concludes session
+## Key Parameters
+
+- **Max Agents:** 5 (enforced by Claude Code background task limit)
+- **Lease Timeout:** 5 minutes (default, configurable)
+  - Tasks claimed longer than this without heartbeat are auto-released
+- **Heartbeat Interval:** 60 seconds (recommended)
+  - Agents should call `heartbeat()` at least this often
+  - Prevents false timeout while working on long tasks
+- **Cleanup Interval:** 60 seconds
+  - Orchestrator automatically runs `cleanupStaleClaims()` to release orphaned tasks
+- **Database:** SQLite (stored at `.omc/state/swarm.db`)
+  - One database per swarm session
+  - Survives agent crashes
+  - Provides ACID guarantees
+
+## Error Handling & Recovery
+
+### Agent Crash
+- Task is claimed but agent stops sending heartbeats
+- After 5 minutes of no heartbeat, cleanupStaleClaims() releases the task
+- Task returns to 'pending' status for another agent to claim
+- Original agent's incomplete work is safely abandoned
+
+### Task Completion Failure
+- Agent calls `completeTask()` but is no longer the owner (was released)
+- The update silently fails (no agent matches in WHERE clause)
+- Agent can detect this by checking return value
+- Agent should log error and continue to next task
+
+### Database Unavailable
+- `startSwarm()` returns false if database initialization fails
+- `claimTask()` returns `{ success: false, reason: 'Database not initialized' }`
+- Check `isSwarmReady()` before proceeding
+
+### All Agents Idle
+- Orchestrator detects via `getActiveAgents() === 0` or `hasPendingWork() === false`
+- Triggers final cleanup and marks swarm as complete
+- Remaining failed tasks are preserved in database
+
+### No Tasks Available
+- `claimTask()` returns success=false with reason 'No pending tasks available'
+- Agent should check `hasPendingWork()` before looping
+- Safe for agent to exit cleanly when no work remains
 
 ## Cancellation
 
@@ -233,42 +416,61 @@ Use unified cancel command:
 This:
 - Stops orchestrator monitoring
 - Signals all background agents to exit
-- Preserves partial progress in swarm-tasks.json
+- Preserves partial progress in database
 - Marks session as "cancelled"
 
 ## Use Cases
 
 ### Fix All Type Errors
 ```
-/swarm 5:executor "fix all TypeScript type errors"
+/oh-my-claudecode:swarm 5:executor "fix all TypeScript type errors"
 ```
 Spawns 5 executors, each claiming and fixing individual files.
 
 ### Implement UI Components
 ```
-/swarm 3:designer "implement Material-UI styling for all components"
+/oh-my-claudecode:swarm 3:designer "implement Material-UI styling for all components"
 ```
 Spawns 3 designers, each styling different component files.
 
 ### Security Audit
 ```
-/swarm 4:security-reviewer "review all API endpoints for vulnerabilities"
+/oh-my-claudecode:swarm 4:security-reviewer "review all API endpoints for vulnerabilities"
 ```
 Spawns 4 security reviewers, each auditing different endpoints.
 
 ### Documentation Sprint
 ```
-/swarm 2:writer "add JSDoc comments to all exported functions"
+/oh-my-claudecode:swarm 2:writer "add JSDoc comments to all exported functions"
 ```
 Spawns 2 writers, each documenting different modules.
 
-## Benefits
+## Benefits of SQLite-Based Implementation
+
+### Atomicity & Safety
+- **Race-Condition Free:** SQLite transactions guarantee only one agent claims each task
+- **No Lost Updates:** ACID compliance means state changes are durable
+- **Orphan Prevention:** Expired claims are automatically released without manual intervention
+
+### Performance
+- **Fast Queries:** Indexed lookups on task status and agent ID
+- **Concurrent Access:** Multiple agents read/write without blocking
+- **Minimal Lock Time:** Transactions are microseconds, not seconds
+
+### Reliability
+- **Crash Recovery:** Database survives agent failures
+- **Automatic Cleanup:** Stale claims don't block progress
+- **Lease-Based:** Time-based expiration prevents indefinite hangs
+
+### Developer Experience
+- **Simple API:** Just `claimTask()`, `completeTask()`, `heartbeat()`
+- **Full Visibility:** Query any task or agent status at any time
+- **Easy Debugging:** SQL queries show exact state without decoding JSON
 
-- **Parallel Execution:** N agents work simultaneously
-- **Auto-Balancing:** Fast agents claim more tasks
-- **Fault Tolerance:** Timeouts prevent deadlocks
-- **Progress Visibility:** Live stats on claimed/done/pending
-- **Scalable:** Works for 10s to 100s of subtasks
+### Scalability
+- **10s to 1000s of Tasks:** SQLite handles easily
+- **Full Task Retention:** Complete history in database for analysis
+- **Extensible Schema:** Add custom columns for task metadata
 
 ## Output
 

package/dist/__tests__/hooks/auto-slash-command/executor.test.d.ts

@@ -0,0 +1,7 @@
+/**
+ * Integration tests for learned skill discovery
+ *
+ * Tests the discoverAllCommands() functionality with project and user-scoped skills
+ */
+export {};
+//# sourceMappingURL=executor.test.d.ts.map

package/dist/__tests__/hooks/auto-slash-command/executor.test.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"executor.test.d.ts","sourceRoot":"","sources":["../../../../src/__tests__/hooks/auto-slash-command/executor.test.ts"],"names":[],"mappings":"AAAA;;;;GAIG"}