WogiFlow

A self-improving AI development workflow that learns from your feedback. Currently supports Claude Code 2.1.33+.

npm install -D wogiflow
# or
bun add -d --trust wogiflow   # --trust is required for bun to run postinstall

npx flow onboard        # Analyze your project
npx flow bridge sync    # Sync to Claude Code

Features

| Feature | Description | |---------|-------------| | Task Gating | Blocks implementation without an active tracked task | | Self-Completing Tasks | /wogi-start runs until all acceptance criteria pass | | Multi-Agent Explore Phase | 5-6 parallel research agents analyze codebase before implementation | | Extensible Registry | Plugin-based component/function/API registries with auto-activation | | AI-Judge Semantic Matching | Detects duplicate components, functions, and APIs using AI similarity | | Framework Discovery | Auto-detects your stack and activates relevant registry plugins | | Consumer Impact Analysis | Maps all code consumers before refactoring to prevent breakage | | Component Registry | Tracks all components in app-map.md — prevents duplicates | | Function Registry | Tracks utility functions in function-map.md — prevents duplicates | | API Registry | Tracks endpoints in api-map.md — prevents duplicates | | Adversarial Code Review | Review agents must find minimum findings or justify clean code | | Git-Verified Claims | Cross-references spec deliverables against actual git diff (opt-in) | | Standards Compliance | Auto-checks naming conventions, security patterns, and project rules | | TDD Mode | Opt-in test-first development (write test → fail → implement → pass) | | Phased Execution | Contract → Skeleton → Core → Edge Cases → Polish | | Hybrid Mode | Claude plans, local LLM executes — configurable token savings | | Peer Review | Multi-model code review for diverse perspectives | | Skills System | Modular add-ons that learn from your sessions | | Research Protocol | Enforces verification before capability claims | | Worktree Isolation | Parallel tasks run in isolated git worktrees | | Parallel Execution | Independent tasks execute concurrently | | Durable Sessions | Suspend/resume tasks across sessions with full context | | Debug Hypothesis | Parallel agents investigate competing theories simultaneously | | Browser Debug/Test | WebMCP-powered browser debugging and test flows | | Decision Tracking | Auditable amendment trail for all project rule changes | | Cross-Artifact Consistency | Validates registries against codebase — detects orphans and phantoms | | Self-Improvement | Learns from corrections — after 3+ similar mistakes, creates permanent rules | | Project Audit | 7-dimension deep analysis: architecture, security, performance, and more | | Eval System | Multi-judge scoring for task output quality assessment |

See the Knowledge Base for detailed documentation on each feature.

Commands

Slash Commands (in Claude Code)

| Category | Commands | Description | |----------|----------|-------------| | Daily | /wogi-morning, /wogi-ready, /wogi-status, /wogi-standup | Briefing, task queue, project overview, standup summary | | Tasks | /wogi-start, /wogi-story, /wogi-bug, /wogi-bulk, /wogi-bulk-loop | Execute, create stories, report bugs, batch process, continuous loop | | Planning | /wogi-feature, /wogi-epics, /wogi-plan, /wogi-deps | Feature/epic/plan management, dependency trees | | Review | /wogi-review, /wogi-review-fix, /wogi-peer-review, /wogi-triage, /wogi-eval | Code review, auto-fix, multi-model, triage, quality eval | | Debug | /wogi-debug-hypothesis, /wogi-debug-browser, /wogi-test-browser, /wogi-trace | Parallel theories, browser debug, test flows, code traces | | Registries | /wogi-map, /wogi-map-add, /wogi-map-scan, /wogi-map-sync, /wogi-map-check, /wogi-map-index | Component registry management | | Rules | /wogi-decide, /wogi-learn, /wogi-rules, /wogi-retrospective | Create rules, promote patterns, session retros | | Research | /wogi-research, /wogi-correction | Zero-trust verification, correction reports | | Context | /wogi-compact, /wogi-context, /wogi-suspend, /wogi-resume | Memory management, task context, suspend/resume | | Capture | /wogi-capture, /wogi-extract-review, /wogi-changelog | Quick capture, transcript extraction, changelogs | | Hybrid | /wogi-hybrid, /wogi-hybrid-setup, /wogi-hybrid-edit, /wogi-hybrid-off, /wogi-hybrid-status | Local LLM integration | | Skills | /wogi-skills, /wogi-skill-learn, /wogi-setup-stack | Skill packages, learning extraction, stack detection | | Session | /wogi-session-end, /wogi-health, /wogi-finalize, /wogi-audit | End session, health check, branch finalization, project audit | | Config | /wogi-config, /wogi-models-setup, /wogi-statusline-setup, /wogi-init, /wogi-onboard, /wogi-register | Configuration, model setup, status line, plugins | | Utilities | /wogi-search, /wogi-guided-edit, /wogi-export, /wogi-import, /wogi-debt, /wogi-roadmap, /wogi-log, /wogi-rescan, /wogi-suggest, /wogi-help | Search, guided edits, export/import, tech debt, roadmap, suggestions |

See Command Reference for full details on every command.

CLI Commands

flow onboard                # Analyze project & set up context
flow bridge sync            # Sync workflow to Claude Code
flow ready                  # Show task queue
flow start <id>             # Start a task
flow status                 # Project overview
flow health                 # Check workflow integrity
flow verify all             # Run all quality gates
flow map-index scan         # Rescan component registry
flow skill-learn            # Extract learnings from session
flow hybrid setup           # Initial setup for local LLM integration
flow hybrid enable          # Enable hybrid mode
flow worktree enable        # Enable worktree isolation
flow parallel check         # Check for parallelizable tasks
flow figma analyze <file>   # Figma-to-code pipeline
flow metrics                # Command success/failure stats

Run flow --help or see CLI docs for the full list.

How Tasks Work

/wogi-start "add logout button"
    │
    ├── Classify request → Route to story creation
    ├── Multi-agent research (5-6 parallel agents)
    ├── Generate spec with acceptance criteria
    ├── Approval gate (for stories/epics)
    │
    │   FOR EACH criterion:
    │   ├── Implement
    │   ├── Verify (lint, typecheck, tests)
    │   └── Mark complete only when passing
    │
    ├── Criteria completion check (re-verify ALL)
    ├── Wiring check (no orphan files)
    ├── Standards compliance check
    ├── Update registries + request-log
    └── Commit

File Structure

.workflow/
├── config.json              # Workflow configuration (500+ options)
├── state/
│   ├── ready.json           # Task queue
│   ├── request-log.md       # Change history
│   ├── decisions.md         # Project rules (auto-learned)
│   ├── feedback-patterns.md # Bug patterns and failure learnings
│   ├── app-map.md           # Component registry
│   ├── function-map.md      # Utility function registry
│   ├── api-map.md           # API endpoint registry
│   ├── session-state.json   # Session persistence
│   └── progress.md          # Session handoff notes
├── specs/                   # Task specifications
├── templates/               # CLAUDE.md generation templates
├── reviews/                 # Code review reports
└── verifications/           # Verification artifacts

.claude/
├── commands/                # 69 slash commands
├── skills/                  # Skill modules (auto-learned)
├── rules/                   # Project coding rules
└── docs/                    # Documentation & knowledge base

Documentation

| Category | Topics | |----------|--------| | Setup & Onboarding | Installation, project analysis, component indexing, framework detection | | Task Execution | Execution pipeline, verification, completion, specifications | | Self-Improvement | Skills, learning, model adapters, pattern promotion | | Memory & Context | Context management, session persistence, PRD management | | Development Tools | Figma integration, code traces, guided edits, MCP integrations | | Safety & Guardrails | Damage control, security scanning, checkpoint/rollback, commit gates | | Configuration | All 500+ configuration options | | Command Reference | Every slash command and CLI command |

License

AGPL-3.0