
Orc
- Updated January 30, 2026
- twofoldtech-dakota/ORC
Orc is an autonomous multi-agent system that transforms goals into tested implementations, organizing work as an Epic/Feature/Story hierarchy with quality gates between stages. Developers use it to drive end-to-end feature delivery with multiple coordinated agents and built-in verification.
Key points
- Autonomous multi-agent system
- Epic/Feature/Story hierarchy
- Quality gates, tested output
Orc by the numbers
- Data as of Jul 7, 2026 (Skillselion catalog sync)
/plugin marketplace add twofoldtech-dakota/ORC/plugin install orc@orc-marketplaceAdd your badge
Show developers this skill is listed on Skillselion. Paste this into your README.
| Last updated | January 30, 2026 |
|---|---|
| Repository | twofoldtech-dakota/ORC ↗ |
What it does
Turn goals into tested implementations via an Epic/Feature/Story multi-agent hierarchy with quality gates.
README.md
ORC (Orchestrator)
Multi-Agent Orchestration System for Autonomous Software Development
ORC is a Claude Code plugin that transforms high-level goals into complete, tested implementations through a structured phased approach with mandatory plan approval.
Quick Start
# Create a plan
/orc:plan "Build REST API with user authentication"
# Review the plan
/orc:show
# Approve the plan
/orc:approve
# Execute
/orc:run
Core Philosophy
Quality > Capability > Developer Experience > Speed > Cost
ORC follows a strict workflow:
- Plan Phase - Decompose goal into Epic → Feature → Story hierarchy
- User Approval - Mandatory review before any implementation
- Execute Phase - Implement with ReAct pattern, validate each story
- Review Phase - Final quality gates and deviation analysis
- Learn Phase - Extract patterns for future use
Key Features
Plan-First Execution
No code is written until you approve the plan. Review exactly what will be built before it happens.
Self-Validation
Every story has explicit acceptance criteria verified with evidence. No GUI required—the system validates itself.
Strict Approach Adherence
The Implementer follows suggested approaches exactly. Any deviation is documented, categorized, and flagged for review.
Autonomous Recovery
Failed stories retry with different approaches (up to 3 attempts). If still failing, they're marked blocked and non-dependent work continues.
Parallel Execution
Independent stories execute concurrently. Dependencies are analyzed at plan time to prevent conflicts.
Pattern Learning
Successful implementations are extracted as patterns. Future similar tasks automatically receive proven approaches.
Commands
Planning
| Command | Description |
|---|---|
/orc:plan <goal> |
Create or append to plan |
/orc:show |
Display plan summary |
/orc:show <epic-id> |
Display epic details |
/orc:show deviations |
Show deviations for review |
/orc:approve |
Approve all pending epics |
/orc:approve <epic-id> |
Approve specific epic |
Execution
| Command | Description |
|---|---|
/orc:run |
Execute all approved epics |
/orc:run <epic-id> |
Execute specific epic |
/orc:next |
Execute next priority epic only |
/orc:stop |
Stop execution gracefully |
/orc:resume |
Resume from last checkpoint |
/orc:retry <story-id> |
Retry a blocked story |
Learning & Utility
| Command | Description |
|---|---|
/orc:patterns |
Show learned patterns |
/orc:learn |
Force pattern extraction |
/orc:status |
Show current state |
/orc:clear |
Clear plan and state |
Plan Hierarchy
Plan
└── Epic (self-contained project milestone)
└── Feature (self-contained capability)
└── Story (atomic task with acceptance criteria)
Rules:
- Epics are self-contained (no cross-epic dependencies)
- Features are self-contained (no cross-feature dependencies)
- Stories can depend on stories within the same feature only
Quality Gates
Story Completion
- ✓ All acceptance criteria verified
- ✓ All unit tests pass
- ✓ Type checking passes
- ✓ No lint errors
- ✓ No new security vulnerabilities
- ✓ Approach compliance verified
Feature Completion
- ✓ All stories completed
- ✓ Integration tests pass
- ✓ Feature-level acceptance criteria verified
Epic Completion
- ✓ All features completed
- ✓ E2E tests pass
- ✓ Final review completed
- ✓ Full security scan passes
Agent System
Core Agents (6)
| Agent | Role |
|---|---|
| Orchestrator | State management, phase control |
| Planner | Goal decomposition |
| Implementer | Code implementation (ReAct) |
| Validator | Test & acceptance verification |
| Reviewer | Final quality gate (Reflexion) |
| Learner | Pattern extraction |
Specialists (11)
Spawned on-demand: Architect, Product Designer, DevOps, Security, Database, Frontend, Backend, Full Stack, QA, Biz Analyst, Content Strategist
Runtime Directory
ORC maintains state in .orc/:
.orc/
├── plan/
│ ├── state.json # Execution state
│ ├── plan.json # Master plan
│ ├── learnings.json # Patterns
│ ├── embeddings.json # Vector search
│ ├── deviations.json # Deviation log
│ └── epics/ # Epic definitions
└── checkpoints/ # Recovery points
Documentation
Example Session
> /orc:plan "Build a blog API with authentication"
📋 Plan Created: Build a blog API with authentication
Epics (2):
E1: User Authentication [3 features, 9 stories]
E2: Blog CRUD API [4 features, 14 stories]
Total: 7 features, 23 stories
Run /orc:approve to proceed
> /orc:approve
✓ Plan approved
> /orc:run
Starting execution...
[E1-F1-S1] Creating User model
├─ REASON: Loading pattern sp_bcrypt_001
├─ ACT: Creating src/models/User.ts
├─ VERIFY: ✓ Tests pass (4/4)
└─ COMPLETE ✓
... (continues) ...
[EXECUTION COMPLETE]
├─ Stories: 23/23 completed
├─ Patterns learned: 5 new
└─ Total time: 4m 32s