issue:execute

Minimal orchestrator that dispatches solution IDs to executors. Each executor receives a complete solution with all tasks and commits once per solution.

Description

The issue:execute command executes queued solutions using DAG-based parallel orchestration. Each executor receives a complete solution with all tasks, executes tasks sequentially, and commits once per solution. Supports optional git worktree isolation for clean workspace management.

Key Features

• DAG-based parallelism: Automatic parallel execution of independent solutions
• Solution-level execution: Each executor handles all tasks in a solution
• Single commit per solution: Clean git history with formatted summaries
• Worktree isolation: Optional isolated workspace for queue execution
• Multiple executors: Codex, Gemini, or Agent support
• Resume capability: Recover from interruptions with existing worktree

Usage

# Execute specific queue (REQUIRED)
/issue:execute --queue QUE-xxx

# Execute in isolated worktree
/issue:execute --queue QUE-xxx --worktree

# Resume in existing worktree
/issue:execute --queue QUE-xxx --worktree /path/to/worktree

# Dry-run (show DAG without executing)
/issue:execute --queue QUE-xxx
# Select: Dry-run mode

Arguments

Argument	Required	Description
--queue <id>	Yes	Queue ID to execute (required)
--worktree	No	Create isolated worktree for execution
--worktree <path>	No	Resume in existing worktree

Executor Selection

Interactive prompt selects:

• Codex (Recommended): Autonomous coding with 2hr timeout
• Gemini: Large context analysis and implementation
• Agent: Claude Code sub-agent for complex tasks

Examples

Execute Queue (Interactive)

/issue:execute --queue QUE-20251227-143000
# Output:
# ## Executing Queue: QUE-20251227-143000
# ## Queue DAG (Solution-Level)
# - Total Solutions: 5
# - Ready: 2
# - Completed: 0
# - Parallel in batch 1: 2
#
# Select executor:
#   [1] Codex (Recommended)
#   [2] Gemini
#   [3] Agent
# Select mode:
#   [1] Execute (Recommended)
#   [2] Dry-run
# Use git worktree?
#   [1] Yes (Recommended)
#   [2] No

Queue ID Not Provided

/issue:execute
# Output:
# Available Queues:
# ID                    Status      Progress    Issues
# -----------------------------------------------------------
# → QUE-20251215-001   active      3/10        ISS-001, ISS-002
#   QUE-20251210-002   active      0/5         ISS-003
#   QUE-20251205-003   completed   8/8         ISS-004
#
# Which queue would you like to execute?
#   [1] QUE-20251215-001 - 3/10 completed, Issues: ISS-001, ISS-002
#   [2] QUE-20251210-002 - 0/5 completed, Issues: ISS-003

Execute with Worktree

/issue:execute --queue QUE-xxx --worktree
# Output:
# Created queue worktree: /repo/.ccw/worktrees/queue-exec-QUE-xxx
# Branch: queue-exec-QUE-xxx
# ### Executing Solutions (DAG batch 1): S-1, S-2
# [S-1] Executor launched (codex, 2hr timeout)
# [S-2] Executor launched (codex, 2hr timeout)
# ✓ S-1 completed: 3 tasks, 1 commit
# ✓ S-2 completed: 2 tasks, 1 commit

Resume Existing Worktree

# Find existing worktrees
git worktree list
# /repo/.ccw/worktrees/queue-exec-QUE-123

# Resume execution
/issue:execute --queue QUE-123 --worktree /repo/.ccw/worktrees/queue-exec-QUE-123
# Output:
# Resuming in existing worktree: /repo/.ccw/worktrees/queue-exec-QUE-123
# Branch: queue-exec-QUE-123
# ### Executing Solutions (DAG batch 2): S-3

Issue Lifecycle Flow

Execution Model

DAG-Based Batching

Batch 1 (Parallel): S-1, S-2  → No file conflicts
Batch 2 (Parallel): S-3, S-4  → No conflicts, waits for Batch 1
Batch 3 (Sequential): S-5     → Depends on S-3

Solution Execution (Within Executor)

ccw issue detail S-1  → Get full solution with all tasks
↓
For each task T1, T2, T3...:
  - Follow implementation steps
  - Run test commands
  - Verify acceptance criteria
↓
After ALL tasks pass:
  git commit -m "feat(scope): summary

  Solution: S-1
  Tasks completed: T1, T2, T3

  Changes:
  - file1: what changed
  - file2: what changed

  Verified: all tests passed"
↓
ccw issue done S-1 --result '{summary, files, commit}'

Executor Dispatch

Codex Executor

ccw cli -p "## Execute Solution: S-1
..." --tool codex --mode write --id exec-S-1
# Timeout: 2 hours (7200000ms)
# Background: true

Gemini Executor

ccw cli -p "## Execute Solution: S-1
..." --tool gemini --mode write --id exec-S-1
# Timeout: 1 hour (3600000ms)
# Background: true

Agent Executor

Task({
  subagent_type: 'code-developer',
  run_in_background: false,
  description: 'Execute solution S-1',
  prompt: '...'  // Full execution prompt
})

Worktree Management

Create New Worktree

# One worktree for entire queue execution
git worktree add .ccw/worktrees/queue-exec-QUE-xxx -b queue-exec-QUE-xxx
# All solutions execute in this isolated workspace
# Main workspace remains untouched

Resume Existing Worktree

# Find interrupted executions
git worktree list
# Output:
# /repo/.ccw/worktrees/queue-exec-QUE-123    abc1234 [queue-exec-QUE-123]

# Resume with worktree path
/issue:execute --queue QUE-123 --worktree /repo/.ccw/worktrees/queue-exec-QUE-123

Worktree Completion

After all batches complete:

# Prompt for merge strategy
Queue complete. What to do with worktree branch "queue-exec-QUE-xxx"?
  [1] Create PR (Recommended)
  [2] Merge to main
  [3] Keep branch

# Create PR
git push -u origin queue-exec-QUE-xxx
gh pr create --title "Queue QUE-xxx" --body "Issue queue execution"
git worktree remove .ccw/worktrees/queue-exec-QUE-xxx

# OR Merge to main
git merge --no-ff queue-exec-QUE-xxx -m "Merge queue QUE-xxx"
git branch -d queue-exec-QUE-xxx
git worktree remove .ccw/worktrees/queue-exec-QUE-xxx

CLI Endpoints

Queue Operations

# List queues
ccw issue queue list --brief --json

# Get DAG
ccw issue queue dag --queue QUE-xxx
# Returns: {parallel_batches: [["S-1","S-2"], ["S-3"]]}

Solution Operations

# Get solution details (READ-ONLY)
ccw issue detail S-1
# Returns: Full solution with all tasks

# Mark solution complete
ccw issue done S-1 --result '{"summary":"...","files_modified":[...],"commit":{...},"tasks_completed":3}'

# Mark solution failed
ccw issue done S-1 --fail --reason '{"task_id":"T2","error_type":"test_failure","message":"..."}'

Commit Message Format

feat(auth): implement OAuth2 login flow

Solution: S-1
Tasks completed: T1, T2, T3

Changes:
- src/auth/oauth.ts: Implemented OAuth2 flow
- src/auth/login.ts: Integrated OAuth with existing login
- tests/auth/oauth.test.ts: Added comprehensive tests

Verified: all tests passed

Commit Types:

• feat: New feature
• fix: Bug fix
• refactor: Code refactoring
• docs: Documentation
• test: Test updates
• chore: Maintenance tasks

Related Commands

• issue:queue - Form execution queue before executing
• issue:plan - Plan solutions before queuing
• ccw issue retry - Reset failed solutions for retry
• ccw issue queue dag - View dependency graph
• ccw issue detail <id> - View solution details

Best Practices

1. Use Codex executor: Best for long-running autonomous work
2. Enable worktree: Keeps main workspace clean during execution
3. Check DAG first: Use dry-run to see execution plan
4. Monitor progress: Executors run in background, check completion
5. Resume on failure: Use existing worktree path to continue
6. Review commits: Each solution produces one formatted commit

Troubleshooting

Error	Cause	Resolution
No queue specified	--queue argument missing	List queues and select one
No ready solutions	Dependencies blocked	Check DAG for blocking issues
Executor timeout	Solution too complex	Break into smaller solutions
Worktree exists	Previous incomplete execution	Resume with --worktree <path>
Partial task failure	Task reports failure	Check ccw issue done --fail output
Git conflicts	Parallel executors touched same files	DAG should prevent this