AI-powered development workflow management system with multi-model support
npm install wogiflowA self-improving AI development workflow that learns from your feedback and works across multiple AI coding CLIs.
``bashInstall
npm install wogiflow
Supported CLIs
WogiFlow works with 6 AI coding CLIs. Use whichever you prefer - the workflow state is shared.
| CLI | Enforcement | Rules File | Min Version | Guide |
|-----|-------------|------------|-------------|-------|
| Claude Code | Hard (hooks) |
CLAUDE.md | 2.1.23+ | Guide |
| Gemini CLI | Hard (hooks) | GEMINI.md | - | Guide |
| Cursor | Mixed | .cursor/rules/wogiflow.mdc | - | Guide |
| OpenCode | Hard (plugins) | AGENTS.md | - | Guide |
| Codex | Soft (rules) | AGENTS.md | - | Guide |
| Kimi | Soft (rules) | AGENTS.md | - | Guide |> Claude Code 2.1.23+ Recommended: Includes critical fixes for per-user temp directory isolation (shared systems), async hook cancellation, and ripgrep timeout reporting. Earlier versions may experience silent search failures.
Enforcement levels:
- Hard: Blocks operations before execution (best protection)
- Mixed: Hard at prompt level, soft after
- Soft: Advisory only (rules in context, no blocking)
$3
Workflow state is stored in
.workflow/state/ - CLI-agnostic. You can:
1. Start a task in Claude Code
2. Continue it in Cursor
3. Finish it in Gemini CLIRun
npx flow bridge sync after switching to regenerate CLI-specific files.---
Core Features
| Feature | Description |
|---------|-------------|
| Task Gating | Blocks implementation without an active task |
| Self-Completing Tasks |
/wogi-start runs until all criteria pass |
| Component Registry | Prevents duplicate components via app-map |
| Post-Edit Validation | Auto-runs lint/typecheck after every edit |
| Request Logging | All changes logged with tags for searchability |
| Hybrid Mode | Claude plans, local LLM executes (20-60% 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 |---
Developer Workflow
$3
`bash
Session start
/wogi-morning # Morning briefing with task recommendations
/wogi-ready # Show available tasks
/wogi-status # Project overviewTask execution
/wogi-start TASK-012 # Start task (self-completing loop)
/wogi-start "add feature" # Or describe what you wantCreation
/wogi-story "Add login" # Create story with acceptance criteria
/wogi-bug "Login fails" # Report a bugSession end
/wogi-review # Code review with 3 parallel agents
/wogi-session-end # Save progress, commit, push
`$3
`
/wogi-start "add logout button"
│
├── [AUTO] Classify request → Implementation
├── [AUTO] Check app-map for existing components
├── [AUTO] Generate acceptance criteria
│
│ FOR EACH edit:
│ ├── [AUTO] Validate file is in scope
│ ├── [AUTO] Run lint/typecheck
│
├── [AUTO] Verify ALL criteria met
├── [AUTO] Update app-map, request-log
└── [AUTO] Commit changes
`$3
| Category | Commands |
|----------|----------|
| Tasks |
/wogi-ready, /wogi-start, /wogi-status, /wogi-deps |
| Create | /wogi-story, /wogi-bug, /wogi-feature |
| Review | /wogi-review, /wogi-peer-review |
| Components | /wogi-map, /wogi-map-add, /wogi-map-scan |
| Session | /wogi-morning, /wogi-session-end, /wogi-health |
| Hybrid | /wogi-hybrid, /wogi-hybrid-setup, /wogi-hybrid-off |
| Utilities | /wogi-search, /wogi-trace, /wogi-config |---
CLI Commands
`bash
Setup
flow onboard # Analyze project, populate state files
flow bridge sync # Sync to current CLITasks
flow ready # Show task queue
flow start # Start task
flow status # Project overviewComponents
flow map-index scan # Rescan codebase
flow map-sync # Compare index with app-mapHybrid Mode
flow hybrid enable # Enable with setup wizard
flow hybrid status # Show configurationSkills
flow skill-learn # Extract learnings from session
flow skill detect # Detect frameworksHealth
flow health # Check workflow integrity
flow verify all # Run all quality gates
`---
File Structure
`
.workflow/
├── config.json # Workflow configuration
├── bridges/ # CLI bridge implementations
├── templates/ # Templates for CLI files
├── docs/cli-guides/ # Per-CLI documentation
└── state/
├── ready.json # Task queue
├── request-log.md # Change history
├── app-map.md # Component registry
├── decisions.md # Project rules
└── progress.md # Session handoff notes.claude/ # Claude Code specific
├── skills/ # Skill modules
├── docs/ # Documentation
└── rules/ # Project rules
CLAUDE.md # Claude Code instructions (generated)
GEMINI.md # Gemini CLI instructions (generated)
AGENTS.md # Codex/Kimi/OpenCode instructions (generated)
`---
Configuration
Main configuration in
.workflow/config.json:`json
{
"enforcement": {
"strictMode": true
},
"hooks": {
"rules": {
"taskGating": { "enabled": true },
"validation": { "enabled": true },
"componentReuse": { "enabled": true }
}
},
"qualityGates": {
"feature": { "require": ["tests", "appMapUpdate", "requestLogEntry"] }
},
"hybrid": {
"enabled": false,
"provider": "ollama"
}
}
``---
Detailed documentation is in the Knowledge Base:
| Category | Topics |
|----------|--------|
| Setup & Onboarding | Installation, onboarding, component indexing |
| Task Execution | Workflow steps, verification, completion |
| Self-Improvement | Skills, learning, model adapters |
| Memory & Context | Context management, session persistence |
| Development Tools | Figma analyzer, code traces, MCP |
| Safety & Guardrails | Damage control, checkpoints, security |
| Configuration | All configuration options |
Each supported CLI has its own guide explaining setup, enforcement, and usage:
- Claude Code Guide
- Gemini CLI Guide
- Cursor Guide
- OpenCode Guide
- Codex Guide
- Kimi Guide
---
WogiFlow learns from your corrections:
1. Correction → You correct the AI's work
2. Fix → AI fixes immediately
3. Learn → AI asks to persist the rule
4. Update → Updates decisions.md / skills / config
5. Track → Logs to feedback-patterns.md
After 3+ similar corrections → promotes to permanent instruction.
---
AGPL-3.0