Unified OS + Application Layer for Claude Code Development
Finish What You Start. Sessions that Last. Features that Ship.
SuperNavigator combines two powerful frameworks into a unified layered architecture:
OS Layer (Navigator): Context engineering, session management, Theory of Mind
- 92% token reduction through smart context loading
- Persistent memory across sessions (nav-profile, nav-marker)
- Loop mode for autonomous task completion
- OpenTelemetry metrics and Grafana dashboards
Application Layer (Superpowers): Development workflows, TDD, collaboration patterns
- Complete software development workflow from design to deployment
- Mandatory TDD with test-first enforcement
- Systematic debugging and code review processes
- Parallel subagent development for complex features
Integration: Both layers communicate via shared .agent/ data store with implicit triggers.
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β Application Layer β
β (Superpowers: Workflows, TDD, Debugging, Collaboration) β
β β
β β’ brainstorming β writing-plans β executing-plans β
β β’ test-driven-development β requesting-code-review β
β β’ systematic-debugging β verification-before-completion β
β β’ using-git-worktrees β finishing-a-development-branch β
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β
.agent/ Data Store
(Implicit Integration Layer)
β
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β OS Layer β
β (Navigator: Context, Memory, Sessions, Metrics) β
β β
β β’ nav-init β nav-start β nav-loop β
β β’ nav-marker β nav-compact (context management) β
β β’ nav-profile β nav-diagnose (Theory of Mind) β
β β’ nav-stats (metrics) β nav-task/nav-sop (docs) β
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
Install as Claude Code Plugin:
claude plugin marketplace add https://github.com/tennyson-mccalla/SuperNavigator
claude plugin install supernavigatorThen restart Claude Code. All 31 skills available immediately.
See Plugin Installation Guide for:
- Detailed setup
- Verification
- Troubleshooting
- Managing the plugin
"Initialize SuperNavigator in this project"
Creates .agent/ structure with both layers enabled.
"Start my SuperNavigator session"
Loads OS layer context + enables Application layer workflows.
Before you code:
nav-startloads minimal context (12k tokens vs 150k)nav-profileapplies your preferencesnav-markerpreserves important checkpoints
While you code:
nav-compactauto-triggers at 85% context usagenav-diagnosecatches collaboration driftnav-loopenables autonomous task completion
After you code:
nav-statsshows token efficiency (92% reduction)nav-task/nav-sopdocument learnings
Design Phase:
brainstormingrefines ideas through Socratic dialoguewriting-planscreates detailed implementation tasks
Implementation Phase:
using-git-worktreescreates isolated workspacetest-driven-developmentenforces RED-GREEN-REFACTORexecuting-plansorsubagent-driven-developmentdrives execution
Quality Phase:
requesting-code-reviewvalidates against specsystematic-debuggingroot-cause analysis for bugsverification-before-completionensures it's actually fixed
Completion Phase:
finishing-a-development-branchhandles merge/PR decisions- Auto-saves markers at workflow boundaries (implicit integration)
Layers communicate automatically:
Trigger Points:
- Before
brainstormingβnav-markersaves checkpoint - After
writing-plansβnav-taskindexes plan - After
requesting-code-reviewβnav-compactif needed - On
finishing-a-development-branchβnav-marker+nav-sopupdate
No manual coordination needed - the layers work together seamlessly.
Core:
- nav-init, nav-start, nav-onboard, nav-loop
Context/Memory:
- nav-profile (Theory of Mind - remembers preferences)
- nav-marker (context checkpoints - 97% compression)
- nav-compact (intelligent context pruning)
- nav-diagnose (quality detection)
Documentation:
- nav-task, nav-sop, nav-skill-creator
Project Management:
- nav-stats (metrics), nav-update-claude, nav-release, nav-upgrade, nav-install-multi-claude
Plugin:
- plugin-slash-command
Design:
- brainstorming, writing-plans
Development:
- test-driven-development, using-git-worktrees, executing-plans
Quality:
- requesting-code-review, receiving-code-review, systematic-debugging
Parallel:
- dispatching-parallel-agents, subagent-driven-development
Advanced:
- verification-before-completion, finishing-a-development-branch, writing-skills, using-superpowers
OS Layer (Navigator):
- Context efficiency as primary concern
- Load what you need, when you need it
- Theory of Mind for human-AI synergy
- Metrics-driven optimization
App Layer (Superpowers):
- Test-Driven Development (write tests first, always)
- Systematic over ad-hoc (process over guessing)
- Complexity reduction (simplicity as primary goal)
- Evidence over claims (verify before declaring success)
Unified:
- Layered architecture enables specialization
- Implicit integration reduces cognitive load
- Natural language throughout (no commands to memorize)
- Composable skills that work together
1. "Start my SuperNavigator session"
β OS layer loads context (12k tokens)
β App layer workflows enabled
2. "Let's design user authentication"
β brainstorming skill activates
β nav-marker checkpoint created automatically
β Design written to docs/plans/ and .agent/system/
3. "Create implementation plan"
β writing-plans skill activates
β Plan written to .agent/tasks/
β nav-task indexes automatically
4. "Set up isolated workspace"
β using-git-worktrees creates branch
5. "Implement with TDD"
β test-driven-development enforces test-first
β Context stays <50% (nav-compact ready)
6. "Request code review"
β requesting-code-review validates work
β nav-compact auto-triggers if needed
7. "Finish this branch"
β finishing-a-development-branch presents options
β nav-marker checkpoint created
β nav-sop suggested for learnings
Not estimates. Verified via OpenTelemetry.
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β SUPERNAVIGATOR EFFICIENCY REPORT β
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
π OS LAYER (Context Management)
ββββββββββββββββββββββββββββββββββββββββββββββββββββ
Tokens saved: 138,000 (92% β)
Session length: 20+ exchanges
Context usage: 35% (excellent)
π APP LAYER (Development Workflows)
ββββββββββββββββββββββββββββββββββββββββββββββββββββ
Tests written first: 100%
Code review pass rate: 87%
Autonomous task completion: 65%
- Architecture Guide - Layered system deep dive
- Getting Started - Installation and first session
- OS Layer Guide - Navigator features
- App Layer Guide - Superpowers workflows
- Implicit Integration - How layers communicate
- Migration from Navigator
- Migration from Superpowers
After initialization, configure in .agent/.nav-config.json:
{
"version": "6.0.0",
"layers": {
"os_layer_enabled": true,
"app_layer_enabled": true
},
"tom_features": {
"verification_checkpoints": true,
"profile_enabled": true,
"diagnose_enabled": true
},
"app_layer_features": {
"tdd_enforced": true,
"systematic_debugging": true
},
"implicit_integration": {
"auto_save_markers": true,
"auto_compact_threshold": 0.85,
"auto_update_nav_tasks": true
}
}Problem: AI coding sessions crash at exchange 5-7. Context fills with unused docs.
Navigator's solution: Context engineeringβload what you need, when you need it.
Superpowers' solution: Structured workflowsβbrainstorming, TDD, systematic debugging.
SuperNavigator's solution: Both. Context management AND structured development. Together.
Result:
- Sessions go 20+ exchanges (from Navigator)
- 100% test-first development (from Superpowers)
- Features actually ship
- Context stays efficient
- No manual integration needed
MIT - See LICENSE
Built on the shoulders of:
- Navigator by Aleks Petrov - https://github.com/alekspetrov/navigator
- Superpowers by Jesse Vincent - https://github.com/obra/superpowers
Unified and extended for Claude Code.
SuperNavigator 6.0.0
- Navigator 5.2.0 (OS Layer)
- Superpowers 4.0.3 (App Layer)
- First unified release: 2026-01-20