Skip to content
spec-parallel logo

spec-parallel

Auto-detect and parallelize Spec Kit tasks

SKILL.md

Full skill instructions

Spec Kit Parallel Executor

I automatically detect when Spec Kit tasks have parallel-eligible items and offer to spawn agents.

Activation

I activate when:

  • /​tasks command completes
  • User explicitly requests spec parallelization
  • .spec/​tasks.md is updated with new tasks

Detection Logic

def should_parallelize(tasks_file):
    tasks = parse_tasks(tasks_file)
    parallel_tasks = [t for t in tasks if "[P]" in t or "(P)" in t]
    return len(parallel_tasks) >= 3

Spec Context

I include all Spec Kit artifacts in each agent's context:

FilePurposeRequired
.spec/​spec.mdFeature specificationYes
.spec/​plan.mdImplementation planRecommended
.spec/​design.mdDesign decisionsOptional
.spec/​tasks.mdTask breakdownYes

Agent Prompt Template

You are a Spec Kit implementation agent.

## Your Task
{task_description}

## Feature Specification
{spec_content}

## Implementation Plan
{plan_content}

## Design Decisions
{design_content}

## Instructions
1. Implement ONLY the assigned task
2. Follow the specification exactly
3. Adhere to the implementation plan
4. Write tests for new functionality
5. Commit with message: "feat(spec): {task_name}"
6. Output TASK_COMPLETE when done

Begin implementation.

Files

  • spawn-tasks.sh - Parse tasks and spawn agents

Integration with Base

This skill extends the base parallel-executor skill:

  • Uses same worktree pattern
  • Same monitoring commands
  • Same merge workflow

Spec Validation

After tasks complete, I can validate against spec:

# Check each worktree's implementation
for wt in ../​project-*; do
    claude -p "Validate implementation in $wt against .spec/​spec.md" --print
done