workflow - Task Automation

workflow provides adversarial quality loops, task tracking, and process improvement suggestions for high-quality software development.

Why workflow?

Adversarial Quality Assurance

Every task goes through adversarial verification - security audits, quality checks, and documentation reviews happen automatically.

Key Features

• 🔄 Adversarial Loops: Automatic quality verification cycles
• 📊 Task Tracking: Monitor subagent spawns and completions
• 💡 Memory Suggestions: Intelligent suggestions for preserving learnings
• 📈 Process Improvement: Identify common oversights and suggest improvements

Installation

# Install from marketplace
claude mcp add-dir https://github.com/standardbeagle/standardbeagle-tools
claude mcp add workflow --source ./plugins/workflow

Available Skills

Skill	Description
setup-workflow	Configure project-specific workflow rules
start-loop	Start adversarial Ralph Wiggum loop
stop-loop	Stop running workflow loop
loop-status	Show current loop status
add-task	Add task to workflow list
review-memories	Manage workflow memories

Usage Examples

Start Quality Loop

# Start adversarial loop for general tasks
/start-loop

# Check loop status
/loop-status

Add Tasks

# Add a new task
/add-task "Implement user authentication"

# Track progress
/loop-status

Review and Save Memories

# Review collected memories
/review-memories

# Before context compaction, suggestions appear automatically

Adversarial Quality Pipeline

The workflow plugin runs tasks through multiple verification phases:

Phase 1: Implementation

• Task executor implements the solution
• Tests are written (TDD encouraged)
• Code is reviewed

Phase 2: Quality Verification

• Independent agent verifies implementation
• Checks against original requirements
• Identifies edge cases and issues

Phase 3: Security Audit

• OWASP Top 10 vulnerability check
• Input validation review
• Authentication/authorization audit

Phase 4: Documentation

• README updates
• API documentation
• Inline comments for complex logic

Hooks

workflow includes comprehensive lifecycle hooks:

Hook	Script	Purpose
SessionStart	session-init.js	Initialize workflow session
SubagentStop (task-executor)	track-loop-iteration.js	Track task completion
SubagentStop (quality-verifier)	track-verification.js	Track verification
SubagentStop (security-auditor)	track-security-audit.js	Track security audit
PostToolUse (Task)	track-subagent-spawn.js	Track subagent spawns
Stop	session-cleanup.js	Archive session
PreCompact	suggest-memories.js	Suggest memory preservation
PreCompact	suggest-process-improvements.js	Suggest improvements

Session Tracking

workflow tracks session metrics in .claude/workflow-session.json:

{
  "session_id": "session-1234567890-abc123",
  "started_at": "2024-01-15T10:30:00Z",
  "workflow_version": "0.1.0",
  "subagent_spawns": 5,
  "subagent_completions": 4,
  "context_barriers_enforced": 4,
  "loops_started": 1,
  "tasks_completed": 3,
  "tasks_failed": 0,
  "verification_completions": 3,
  "security_audits": 2
}

Memory Preservation

Before context compaction, workflow suggests memories to save:

Memory Categories

1. Workflow Patterns: Effective strategies discovered
2. Technical Decisions: Architecture choices and rationale
3. Code Patterns: Naming conventions, file organization
4. Verification Strategies: What catches bugs effectively
5. Lessons Learned: What went wrong and how to prevent it

Example Memory Suggestion

title: "Auth tasks require rate limiting verification"
category: workflow_pattern
scope: project
content: |
  When implementing authentication endpoints, always include rate limiting
  in the verification phase.

  Checklist:
  - [ ] Rate limit login endpoint (5 attempts/min/IP)
  - [ ] Rate limit password reset (3 attempts/hour/email)
  - [ ] Test with concurrent requests
  - [ ] Verify 429 responses with Retry-After header
tags: [authentication, security, rate-limiting]
confidence: high

Process Improvement Suggestions

workflow identifies common oversights:

1. Testing Coverage: Edge cases, mutation testing, TDD
2. Documentation: README, API docs, examples
3. Root Cause Analysis: Fix causes, not symptoms
4. Security: Input validation, OWASP checks
5. Error Handling: Helpful messages, logging
6. Performance: N+1 queries, caching
7. Backward Compatibility: Breaking changes, migration paths
8. Code Review: Adversarial, thorough
9. Dependencies: Security vulnerabilities, licenses
10. Deployment: Rollback plans, monitoring

Agents

task-executor

Executes workflow tasks with clean context and adversarial verification.

quality-verifier

Independent adversarial verification of implementations.

security-auditor

OWASP-focused security audit of code changes.

Configuration

Project Rules

Configure project-specific rules in .claude/workflow-rules.yaml:

roles:
  - name: security-reviewer
    triggers:
      - file_pattern: "src/auth/**"
      - file_pattern: "src/api/**"
    actions:
      - run_security_audit: true

  - name: test-coverage-checker
    triggers:
      - file_pattern: "src/**/*.ts"
    actions:
      - require_tests: true
      - minimum_coverage: 80

Troubleshooting

Loop Not Starting

# Check workflow session exists
ls .claude/workflow-session.json

# Initialize manually
node plugins/workflow/scripts/session-init.js

Memories Not Saving

1. Check PreCompact hook is configured
2. Verify .claude directory exists
3. Ensure sufficient session activity

Related Resources

• dartai Plugin - Dart task management
• agnt Plugin - Browser debugging
• lci Plugin - Code intelligence

Version History

Version	Changes
0.1.0	Cross-platform Node.js hooks, memory suggestions