Agent Skills: pdca-tracker

PDCA cycle tracking skill for plan-do-check-act improvement management.

continuous-improvementID: a5c-ai/babysitter/pdca-tracker

Install this agent skill to your local

pnpm dlx add-skill https://github.com/a5c-ai/babysitter/tree/HEAD/plugins/babysitter/skills/babysit/process/specializations/domains/science/industrial-engineering/skills/pdca-tracker

Skill Files

Browse the full folder contents for pdca-tracker.

Download Skill

Loading file tree…

plugins/babysitter/skills/babysit/process/specializations/domains/science/industrial-engineering/skills/pdca-tracker/SKILL.md

Skill Metadata

Name
pdca-tracker
Description
PDCA cycle tracking skill for plan-do-check-act improvement management.

pdca-tracker

You are pdca-tracker - a specialized skill for tracking PDCA (Plan-Do-Check-Act) cycles and improvement management.

Overview

This skill enables AI-powered PDCA tracking including:

  • PDCA cycle setup and management
  • Hypothesis development
  • Experiment planning
  • Results verification
  • Standard work updates
  • Cycle iteration tracking
  • Learning documentation
  • Multi-project portfolio view

Capabilities

1. PDCA Cycle Setup

from dataclasses import dataclass
from typing import List, Dict, Optional
from datetime import datetime, timedelta
from enum import Enum
import uuid

class PDCAPhase(Enum):
    PLAN = "plan"
    DO = "do"
    CHECK = "check"
    ACT = "act"

@dataclass
class PDCACycle:
    id: str
    title: str
    owner: str
    start_date: datetime
    current_phase: PDCAPhase
    iteration: int = 1

def create_pdca_cycle(title: str, owner: str, hypothesis: str,
                     success_criteria: Dict):
    """
    Create new PDCA cycle

    hypothesis: What we believe will happen
    success_criteria: Measurable criteria for success
    """
    cycle_id = str(uuid.uuid4())[:8]

    cycle = {
        "id": cycle_id,
        "title": title,
        "owner": owner,
        "created_date": datetime.now().strftime("%Y-%m-%d"),
        "iteration": 1,
        "current_phase": "PLAN",
        "phases": {
            "PLAN": {
                "status": "in_progress",
                "hypothesis": hypothesis,
                "success_criteria": success_criteria,
                "planned_actions": [],
                "resources_needed": [],
                "timeline": None,
                "completed_date": None
            },
            "DO": {
                "status": "not_started",
                "actions_taken": [],
                "observations": [],
                "data_collected": [],
                "issues_encountered": [],
                "completed_date": None
            },
            "CHECK": {
                "status": "not_started",
                "results": {},
                "hypothesis_validated": None,
                "learnings": [],
                "completed_date": None
            },
            "ACT": {
                "status": "not_started",
                "decision": None,  # standardize, adjust, abandon
                "standard_work_updates": [],
                "next_cycle_needed": None,
                "completed_date": None
            }
        },
        "history": []
    }

    return cycle

2. Plan Phase Management

def develop_plan(cycle: Dict, plan_details: Dict):
    """
    Develop the Plan phase

    plan_details: {
        'actions': [{'description': str, 'owner': str, 'due_date': str}],
        'timeline': {'start': str, 'end': str},
        'resources': [str],
        'risks': [str]
    }
    """
    cycle['phases']['PLAN']['planned_actions'] = plan_details.get('actions', [])
    cycle['phases']['PLAN']['timeline'] = plan_details.get('timeline')
    cycle['phases']['PLAN']['resources_needed'] = plan_details.get('resources', [])
    cycle['phases']['PLAN']['risks'] = plan_details.get('risks', [])

    # Validate plan completeness
    validation = validate_plan(cycle['phases']['PLAN'])

    if validation['is_complete']:
        cycle['phases']['PLAN']['status'] = 'complete'
        cycle['phases']['PLAN']['completed_date'] = datetime.now().strftime("%Y-%m-%d")
        cycle['current_phase'] = 'DO'
        cycle['phases']['DO']['status'] = 'in_progress'

        # Log transition
        cycle['history'].append({
            'timestamp': datetime.now().isoformat(),
            'event': 'phase_transition',
            'from': 'PLAN',
            'to': 'DO'
        })

    return {
        'cycle': cycle,
        'validation': validation
    }

def validate_plan(plan: Dict):
    """Validate plan completeness"""
    issues = []

    if not plan.get('hypothesis'):
        issues.append("Missing hypothesis")
    if not plan.get('success_criteria'):
        issues.append("Missing success criteria")
    if not plan.get('planned_actions'):
        issues.append("No actions planned")
    if not plan.get('timeline'):
        issues.append("No timeline defined")

    return {
        'is_complete': len(issues) == 0,
        'issues': issues
    }

3. Do Phase Tracking

def track_do_phase(cycle: Dict, execution_data: Dict):
    """
    Track execution in Do phase

    execution_data: {
        'action_id': str,
        'status': str,
        'observations': [str],
        'data_points': [{'metric': str, 'value': float, 'timestamp': str}],
        'issues': [str]
    }
    """
    do_phase = cycle['phases']['DO']

    # Update action status
    for action in cycle['phases']['PLAN']['planned_actions']:
        if action.get('id') == execution_data.get('action_id'):
            action['status'] = execution_data['status']
            action['actual_completion'] = datetime.now().strftime("%Y-%m-%d")

    # Record observations
    if execution_data.get('observations'):
        do_phase['observations'].extend(execution_data['observations'])

    # Collect data
    if execution_data.get('data_points'):
        do_phase['data_collected'].extend(execution_data['data_points'])

    # Record issues
    if execution_data.get('issues'):
        do_phase['issues_encountered'].extend(execution_data['issues'])

    # Check if Do phase is complete
    planned_actions = cycle['phases']['PLAN']['planned_actions']
    completed = sum(1 for a in planned_actions if a.get('status') == 'complete')

    if completed == len(planned_actions):
        do_phase['status'] = 'complete'
        do_phase['completed_date'] = datetime.now().strftime("%Y-%m-%d")
        cycle['current_phase'] = 'CHECK'
        cycle['phases']['CHECK']['status'] = 'in_progress'

        cycle['history'].append({
            'timestamp': datetime.now().isoformat(),
            'event': 'phase_transition',
            'from': 'DO',
            'to': 'CHECK'
        })

    return {
        'cycle': cycle,
        'do_phase_progress': {
            'actions_completed': completed,
            'actions_total': len(planned_actions),
            'data_points_collected': len(do_phase['data_collected']),
            'issues_count': len(do_phase['issues_encountered'])
        }
    }

4. Check Phase Analysis

import numpy as np

def analyze_check_phase(cycle: Dict):
    """
    Analyze results in Check phase
    """
    check_phase = cycle['phases']['CHECK']
    plan_phase = cycle['phases']['PLAN']
    do_phase = cycle['phases']['DO']

    results = {}

    # Compare results to success criteria
    success_criteria = plan_phase['success_criteria']
    data_collected = do_phase['data_collected']

    criteria_results = []
    for criterion, target in success_criteria.items():
        # Get data for this metric
        metric_data = [d['value'] for d in data_collected if d['metric'] == criterion]

        if metric_data:
            actual = np.mean(metric_data)
            met = (actual >= target if isinstance(target, (int, float))
                   else str(actual) == str(target))

            criteria_results.append({
                'criterion': criterion,
                'target': target,
                'actual': round(actual, 2) if isinstance(actual, float) else actual,
                'met': met
            })

    # Validate hypothesis
    criteria_met = sum(1 for c in criteria_results if c['met'])
    total_criteria = len(criteria_results)

    hypothesis_validated = criteria_met == total_criteria if total_criteria > 0 else None

    check_phase['results'] = {
        'criteria_results': criteria_results,
        'criteria_met': criteria_met,
        'total_criteria': total_criteria
    }
    check_phase['hypothesis_validated'] = hypothesis_validated

    # Generate learnings
    learnings = generate_learnings(criteria_results, do_phase['observations'],
                                   do_phase['issues_encountered'])
    check_phase['learnings'] = learnings

    return {
        'cycle': cycle,
        'analysis': {
            'hypothesis_validated': hypothesis_validated,
            'success_rate': round(criteria_met / total_criteria * 100, 1) if total_criteria > 0 else 0,
            'criteria_results': criteria_results,
            'learnings': learnings
        }
    }

def generate_learnings(criteria_results, observations, issues):
    """Generate learnings from check phase"""
    learnings = []

    # From criteria results
    for cr in criteria_results:
        if cr['met']:
            learnings.append(f"SUCCESS: {cr['criterion']} achieved target")
        else:
            learnings.append(f"MISS: {cr['criterion']} did not meet target - investigate root cause")

    # From issues
    if issues:
        learnings.append(f"ISSUES: {len(issues)} issues encountered during execution")

    return learnings

5. Act Phase Decision

def complete_act_phase(cycle: Dict, decision: str, next_steps: Dict):
    """
    Complete Act phase with decision

    decision: 'standardize', 'adjust', 'abandon'
    next_steps: {
        'standard_work_updates': [str],
        'next_cycle_hypothesis': str,  # if adjust
        'reason_for_abandonment': str   # if abandon
    }
    """
    act_phase = cycle['phases']['ACT']

    act_phase['decision'] = decision

    if decision == 'standardize':
        act_phase['standard_work_updates'] = next_steps.get('standard_work_updates', [])
        act_phase['next_cycle_needed'] = False

        # Mark cycle complete
        act_phase['status'] = 'complete'
        act_phase['completed_date'] = datetime.now().strftime("%Y-%m-%d")
        cycle['status'] = 'completed'

        cycle['history'].append({
            'timestamp': datetime.now().isoformat(),
            'event': 'cycle_complete',
            'decision': 'standardize',
            'outcome': 'success'
        })

    elif decision == 'adjust':
        act_phase['next_cycle_needed'] = True
        act_phase['next_hypothesis'] = next_steps.get('next_cycle_hypothesis')
        act_phase['adjustments'] = next_steps.get('adjustments', [])

        # Prepare next iteration
        cycle['iteration'] += 1
        new_cycle = prepare_next_iteration(cycle)

        cycle['history'].append({
            'timestamp': datetime.now().isoformat(),
            'event': 'iteration_start',
            'iteration': cycle['iteration'],
            'hypothesis': act_phase['next_hypothesis']
        })

        return {'cycle': cycle, 'next_iteration': new_cycle}

    elif decision == 'abandon':
        act_phase['reason_for_abandonment'] = next_steps.get('reason_for_abandonment')
        act_phase['next_cycle_needed'] = False

        act_phase['status'] = 'complete'
        cycle['status'] = 'abandoned'

        cycle['history'].append({
            'timestamp': datetime.now().isoformat(),
            'event': 'cycle_abandoned',
            'reason': act_phase['reason_for_abandonment']
        })

    return {'cycle': cycle}

def prepare_next_iteration(current_cycle: Dict):
    """Prepare next PDCA iteration"""
    return {
        'iteration': current_cycle['iteration'],
        'hypothesis': current_cycle['phases']['ACT'].get('next_hypothesis'),
        'learnings_from_previous': current_cycle['phases']['CHECK']['learnings'],
        'adjustments': current_cycle['phases']['ACT'].get('adjustments', [])
    }

6. PDCA Portfolio View

def get_portfolio_status(cycles: List[Dict]):
    """
    Get portfolio view of all PDCA cycles
    """
    summary = {
        'total_cycles': len(cycles),
        'by_phase': {phase.value: 0 for phase in PDCAPhase},
        'by_status': {'active': 0, 'completed': 0, 'abandoned': 0},
        'iterations': [],
        'cycle_details': []
    }

    for cycle in cycles:
        # Count by phase
        current_phase = cycle.get('current_phase', 'PLAN')
        summary['by_phase'][current_phase.lower()] += 1

        # Count by status
        status = cycle.get('status', 'active')
        summary['by_status'][status] += 1

        # Track iterations
        summary['iterations'].append(cycle.get('iteration', 1))

        # Cycle summary
        summary['cycle_details'].append({
            'id': cycle['id'],
            'title': cycle['title'],
            'owner': cycle['owner'],
            'phase': current_phase,
            'iteration': cycle.get('iteration', 1),
            'status': status,
            'hypothesis_validated': cycle['phases']['CHECK'].get('hypothesis_validated')
        })

    # Aggregate stats
    summary['avg_iterations'] = round(np.mean(summary['iterations']), 1) if summary['iterations'] else 0
    summary['success_rate'] = round(
        sum(1 for c in cycles if c['phases']['CHECK'].get('hypothesis_validated') == True) /
        len(cycles) * 100, 1
    ) if cycles else 0

    return summary

7. Learning Documentation

def document_learnings(cycle: Dict):
    """
    Create comprehensive learning document from PDCA cycle
    """
    learning_doc = {
        'cycle_id': cycle['id'],
        'title': cycle['title'],
        'date_completed': cycle['phases']['ACT'].get('completed_date'),
        'iterations': cycle.get('iteration', 1),
        'sections': {
            'hypothesis': cycle['phases']['PLAN']['hypothesis'],
            'what_we_tried': [],
            'what_happened': [],
            'what_we_learned': [],
            'what_changed': [],
            'recommendations': []
        }
    }

    # What we tried
    for action in cycle['phases']['PLAN']['planned_actions']:
        learning_doc['sections']['what_we_tried'].append(action['description'])

    # What happened
    for obs in cycle['phases']['DO']['observations']:
        learning_doc['sections']['what_happened'].append(obs)

    for issue in cycle['phases']['DO']['issues_encountered']:
        learning_doc['sections']['what_happened'].append(f"Issue: {issue}")

    # What we learned
    learning_doc['sections']['what_we_learned'] = cycle['phases']['CHECK']['learnings']

    # What changed
    if cycle['phases']['ACT']['decision'] == 'standardize':
        learning_doc['sections']['what_changed'] = cycle['phases']['ACT']['standard_work_updates']
    elif cycle['phases']['ACT']['decision'] == 'adjust':
        learning_doc['sections']['what_changed'] = [
            f"Adjusted hypothesis: {cycle['phases']['ACT'].get('next_hypothesis')}"
        ]

    # Recommendations
    learning_doc['sections']['recommendations'] = generate_recommendations(cycle)

    return learning_doc

def generate_recommendations(cycle: Dict):
    """Generate recommendations based on cycle outcome"""
    recommendations = []

    if cycle['phases']['CHECK'].get('hypothesis_validated'):
        recommendations.append("Document and share successful approach")
        recommendations.append("Consider scaling to other areas")
    else:
        recommendations.append("Review root cause analysis depth")
        if cycle.get('iteration', 1) >= 3:
            recommendations.append("Consider different approach or escalation")

    return recommendations

Process Integration

This skill integrates with the following processes:

  • continuous-improvement-program.js
  • a3-problem-solving-project.js
  • kaizen-event-execution.js

Output Format

{
  "pdca_cycle": {
    "id": "abc12345",
    "title": "Reduce Setup Time",
    "current_phase": "CHECK",
    "iteration": 2
  },
  "plan": {
    "hypothesis": "Standardized tool staging will reduce setup 25%",
    "success_criteria": {"setup_time_minutes": 15}
  },
  "check": {
    "hypothesis_validated": false,
    "actual": 18,
    "gap": 3
  },
  "act": {
    "decision": "adjust",
    "next_hypothesis": "Add visual tool board"
  }
}

Best Practices

  1. Start with hypothesis - Clear, testable statement
  2. Define success criteria - Measurable before starting
  3. Small experiments - Test quickly, learn fast
  4. Document everything - Learning is the product
  5. Iterate deliberately - Each cycle builds on previous
  6. Share learnings - Others benefit from your experiments

Constraints

  • Requires discipline to follow process
  • Not for emergencies requiring immediate action
  • Data collection takes time
  • Multiple iterations may be needed