Task Processor Parallel

• Parallel Execution: Spawn multiple subagents to work on independent tasks simultaneously
• Context Isolation: Each subagent starts with a clean context window
• PR Automation: Automated gap analysis and PR comments after each phase

Parallelization Strategy

Level 1: Phase-Level Parallelization

Launch multiple phases in parallel when no dependencies exist. Use run_shell_command to spawn background Gemini CLI instances:

# Batch 1: Independent phases (run in parallel)
# Note: `-p/--prompt` is deprecated in newer Gemini CLI versions; prefer the positional prompt.
gemini --approval-mode=yolo "Implement Phase 1..." < /dev/null &
gemini --approval-mode=yolo "Implement Phase 5..." < /dev/null &
gemini --approval-mode=yolo "Implement Phase 8..." < /dev/null &

Level 2: Sub-Task Parallelization

Within a single phase, parallelize independent sub-tasks using the same pattern. Ensure each subagent has a specific Task ID to prevent overlap.

Execution Protocol

1. Analyze Dependencies

[... existing dependency analysis logic ...]

2. Create Execution Batches

[... existing batch logic ...]

3. Spawn Subagents

Using run_shell_command for background execution:

To launch multiple phases in parallel, use the shell delegation pattern:

gemini --approval-mode=yolo "$(cat <<'EOF'
You are the tdd-developer. Your Task ID is 1.0.
Implement Phase 1: Comparison.
Work through tasks 1.1 through 1.7 from tasks/[task-file].md.
Follow test-driven development and commit after each task.
EOF
)" < /dev/null &

gemini --approval-mode=yolo "$(cat <<'EOF'
You are the tdd-developer. Your Task ID is 5.0.
Implement Phase 5: Performance...
EOF
)" < /dev/null &

4. Monitor Progress

Check Subagent Status: Use standard shell commands to monitor background processes or check the filesystem for updates (e.g., new commits or PRs).

# Check if gemini processes are running
ps aux | grep gemini

5. Collect Results

When a subagent completes, it notifies the orchestrator:

// Subagent 1 completes Phase 1
Result: {
  phase: 1,
  pr_number: 16,
  files_changed: 6,
  tests_added: 50,
  gap_analysis: "docs/PR_16_GAP_ANALYSIS.md",
  status: "ready_for_review"
}

// Orchestrator receives notification
// Triggers next dependent phase (Phase 2)

6. Automated PR Workflow (Per Subagent)

Each subagent follows this protocol when its phase completes:

1. Run Tests: Full test suite for the phase
2. Create Branch: git checkout -b feat/phase-X-name
3. Push: git push -u origin feat/phase-X-name
4. Create PR: Using gh pr create
5. Wait for CI: 30-second delay
6. Check Status: gh pr view + gh pr checks
7. Generate Gap Analysis: Create docs/PR_X_GAP_ANALYSIS.md
8. Push Gap Analysis: Commit and push
9. Add PR Comment: Detailed changes summary
10. Notify Orchestrator: Return results

Example: Full Autonomous Run

# User starts the processor