agent-11
Health Gecti
- License — License: MIT
- Description — Repository has a description
- Active repo — Last push 0 days ago
- Community trust — 10 GitHub stars
Code Uyari
- fs module — File system access in .github/workflows/validate-agents.yml
Permissions Gecti
- Permissions — No dangerous permissions requested
Bu listing icin henuz AI raporu yok.
Opensource AI development agent suite for solo founders. Deploy 11 specialized agents in Claude Code with one command. https://buymeacoffee.com/jamiewatters
AGENT-11™
Your Elite AI Development Squad
One Founder. Eleven Specialists. Unlimited Potential.
If AGENT-11 has saved you time or helped you ship faster, consider ⭐ starring the repo or buying me a coffee ☕ to support development!
🆕 First Time? · 🚀 Quick Start · 🎯 Workflows · 📖 Docs · 💬 Support
What is AGENT-11?
AGENT-11 deploys 11 specialized AI agents to your project, orchestrating them through proven workflows to build production-ready software. Think of it as hiring an elite development team - except they're AI agents working together seamlessly.
Instead of prompting Claude Code for every task, AGENT-11 provides:
- 🎯 Specialized Agents - Each with domain expertise (requirements, architecture, coding, testing, design, etc.)
- 🎖️ Coordinated Workflows - 20 pre-built missions that orchestrate multiple agents automatically
- 🧠 Persistent Context - Knowledge that survives across sessions and agent handoffs
- ✅ Quality Assurance - Built-in testing, verification, and security protocols
How AGENT-11 Works with BOS-AI
AGENT-11 is the technical execution arm of the BOS-AI ecosystem. BOS-AI handles strategic planning (30 business agents) → AGENT-11 handles software development (11 technical agents).
You can use AGENT-11 standalone without BOS-AI. → Complete BOS-AI Integration Guide
🆕 First Time? Start Here
Complete setup from scratch in 5 minutes. If you already have a project with git initialized, skip to Quick Start.
Prerequisites
- Claude Code installed and running (Get Claude Code)
- GitHub CLI for repo creation:
brew install gh(macOS) or GitHub CLI install - BOS-AI documents (optional): PRD, Vision, Brand docs from BOS-AI
Step-by-Step: New Project Setup
# 1. Create your project directory
mkdir my-project && cd my-project
# 2. Initialize git
git init
# 3. Create GitHub repository (private recommended)
gh repo create my-project --private --source=. --push
# 4. Install AGENT-11
curl -sSL https://raw.githubusercontent.com/TheWayWithin/agent-11/main/project/deployment/scripts/install.sh | bash -s full
# 5. Restart Claude Code
/exit
claude
After Installation: Choose Your Path
Path A: Using BOS-AI Documents (recommended for new products)
# Copy your BOS-AI documents
mkdir -p documents/foundations
cp ~/path/to/bos-ai-docs/*.md documents/foundations/
# Run the full workflow
/foundations init # Process docs into structured YAML
/architect # Design system architecture (select "Engaged Mode")
/bootstrap # Generate project plan (select "Engaged Mode")
/plan status # Verify everything is ready
/coord continue # Start autonomous execution
Path B: Starting from Scratch (no BOS-AI docs)
# Create a simple vision document
echo "# My Product\n\nDescription of what I want to build..." > vision.md
# Initialize with dev-setup mission
/coord dev-setup vision.md
# Or jump straight to MVP
/coord mvp vision.md
Path C: Existing Codebase (adding AGENT-11 to existing project)
# Analyze existing project and create docs
/coord dev-alignment
→ Detailed setup guide | → Troubleshooting
The Complete Workflow: Idea → Shipped Product
BOS-AI Documents → /foundations init → /architect → /bootstrap → /coord continue → 🚀 Shipped
| Step | Command | What It Does | Mode Recommendation | Time |
|---|---|---|---|---|
| 1. Process Docs | /foundations init |
Extracts PRD, Vision, ICP, Brand into structured YAML | N/A (automatic) | 2-5 min |
| 2. Design System | /architect |
Interactive architecture decisions → creates architecture.md | Engaged Mode (first time) | 10-15 min |
| 3. Generate Plan | /bootstrap |
Generates phased project-plan.md with tasks | Engaged Mode (first time) | 1-5 min |
| 4. Check Status | /plan status |
Shows current phase, progress, blockers | N/A | Instant |
| 5. Execute | /coord continue |
Autonomous execution until blocked or phase complete | N/A | Hours-Days |
Step-by-Step Walkthrough
Step 1: /foundations init - Process your BOS-AI documents# First, copy your BOS-AI documents to the foundations directory
mkdir -p documents/foundations
cp ~/Documents/BOS-AI/*.md documents/foundations/
# Then run the extraction
/foundations init
What happens: Scans documents/foundations/ and extracts structured YAML from each document type (PRD → features, tech stack; Vision → goals, mission; Brand → colors, typography, etc.)
Output: Creates .context/structured/*.yaml files that agents can parse directly
When to re-run: After editing any foundation document, run /foundations refresh to sync changes
→ Full /foundations documentation
Step 2: /architect - Design your system architecture/architect # Interactive mode selection (recommended)
/architect --mode engaged # Skip selection, go straight to engaged
/architect --mode auto # Use PRD defaults (experienced users)
Mode Selection:
- Engaged Mode (recommended for new projects): Walks you through 7 decision areas with explanations and trade-offs. Takes 10-15 minutes but prevents costly mistakes.
- Auto Mode: Uses tech stack from PRD + sensible defaults. Best for regenerating after minor changes or when you know exactly what you want.
The 7 Decision Areas (Engaged Mode):
- Application Architecture: Monolith vs modular vs microservices
- Frontend Stack: Framework version, rendering strategy, styling, component library
- Backend & Database: Database config, multi-tenancy strategy, API layer approach
- Authentication: Provider selection, auth methods, session strategy, user roles
- External Integrations: Payments (billing model, webhooks), AI models, email service
- Infrastructure: Frontend/backend hosting, CI/CD pipeline, environments
- Security & Observability: API security, error tracking, analytics, logging
Output: Creates comprehensive architecture.md (~400 lines) with system diagrams, decision log, and integration patterns
Prerequisites: Must run /foundations init first
→ Full /architect documentation
Step 3: /bootstrap - Generate your project plan/bootstrap # Interactive mode selection (recommended)
/bootstrap --mode engaged # Review PRD assumptions with checkpoints
/bootstrap --mode auto # Generate immediately (no questions)
/bootstrap --mode preview # Preview without writing files
Mode Selection:
- Engaged Mode (recommended for new projects): Reviews 5 checkpoints to validate PRD assumptions before generating. Catches misunderstandings early.
- Auto Mode: Trusts your PRD completely and generates immediately. Best for validated PRDs or regeneration.
- Preview Mode: Shows what would be generated without creating files.
The 5 Checkpoints (Engaged Mode):
- Foundation Summary: Verify product info and P0 features extracted correctly
- Tech Stack: Confirm framework versions, integrations, AI model selections
- Priority Validation: Review P0/P1/P2 feature assignments
- Phase Structure: Approve phase breakdown and task distribution
- Final Confirmation: Review all decisions before generating
Output: Creates phased project-plan.md with:
- Phase 1: Fully detailed with tasks, acceptance criteria, and quality gates
- Phase 2+: Outlined with key milestones (rolling wave planning)
Prerequisites: Must run /foundations init first
→ Full /bootstrap documentation
Step 4: /plan status - Check progress/plan status # Overview of current state
/plan phase 1 # Deep dive into specific phase
/plan next # What's coming up
What it shows: Current phase, completed tasks, blockers, next actions, quality gate status
Step 5: /coord continue - Autonomous execution/coord continue # Execute until blocked or phase complete
What happens: The Coordinator reads project-plan.md, identifies the next task, delegates to the appropriate specialist (Developer, Tester, etc.), verifies completion, and moves to the next task.
When it stops: At phase boundaries, when blocked by a question, or when quality gates need verification.
After phase complete:
/coord complete phase 1 # Generate next phase context
/clear # Clear conversation context
/coord continue # Resume from project-plan.md
→ Full mission orchestration documentation
Quick Start Command Sequence
# 1. Copy your BOS-AI docs to the project
cp ~/Documents/BOS-AI/*.md ./documents/foundations/
# 2. Process foundation documents
/foundations init
# 3. Design architecture (FIRST TIME: select "Engaged Mode" when prompted)
/architect
# 4. Generate project plan (FIRST TIME: select "Engaged Mode" when prompted)
/bootstrap
# 5. Review what was created
/plan status
# 6. Start building (autonomous)
/coord continue
Skip BOS-AI? Start directly with /coord dev-setup vision.md or /coord mvp requirements.md
🆕 Current Version: v5.2.0
Latest: Dynamic MCP tool loading with 93% token reduction (51K → 3.3K). No more manual profile switching.
→ Full Release History - Complete sprint-by-sprint changelog
Is AGENT-11 Right for You?
✅ Perfect For:
- Solo founders building products without technical teams
- Developers who want structured AI assistance with quality assurance
- Technical projects requiring multiple specialties (frontend, backend, DevOps, testing)
- Long-running projects needing context preservation across sessions
❌ Not Ideal For:
- Simple scripts or one-off tasks (use Claude Code directly)
- Projects requiring constant human oversight (AGENT-11 is autonomous)
- Non-technical work (use BOS-AI for business operations)
What You Can Build
The Ultimate Proof: AGENT-11 Built by AGENT-11
The deployment system you just used was built by AGENT-11 itself—from concept to working system in under 1 day. Complete automated installation, agent orchestration, configuration management, and error handling. 98% success rate, <1 second deployment.
If AGENT-11 can build itself, it can build anything.
Your projects can include:
- SaaS applications with AI analysis, authentication, payment processing
- Marketplace platforms with OAuth, reviews, community features
- Web applications with Next.js, React, Tailwind CSS, modern stacks
- API services with REST/GraphQL endpoints, third-party integrations
- Business tools with analytics, automation, data pipelines
Time to MVP: 1-3 days with /coord mvp mission
→ See 7+ Real Production Projects
Why AGENT-11 Works
The SaaS Boilerplate Killer: Why pay $300-500 for a static starter kit that still needs customization? AGENT-11 builds your entire application from scratch—authentication, payments, dashboards, APIs—all tailored to your exact requirements. No boilerplate limitations. No template constraints. Just your vision, built to spec.
Proven Performance (v2.0 Results):
- 100% file persistence - Architectural solution eliminates silent failures completely
- 39% effectiveness improvement - Extended thinking + self-verification for better decisions
- 84% token reduction - Context editing + memory optimization enables 30+ hour missions
- 37.5% faster missions - Zero rework from file persistence failures
- Time to MVP: Traditional team (3-6 months) → AGENT-11 (2-4 weeks)
- Cost per Feature: Traditional ($10-50k) → AGENT-11 (<$500)
→ Complete Performance Metrics
Your Squad
Core Squad (4 agents):
- 🎯 Strategist - Product vision and requirements analysis
- 💻 Developer - Code implementation at light speed
- ✅ Tester - Zero bugs reach production (SENTINEL Mode)
- 🚀 Operator - Deploy with confidence
Specialist Squad (7 agents):
- 🏗️ Architect - Bulletproof technical decisions
- 🎨 Designer - Interfaces that convert (RECON Protocol)
- 📚 Documenter - Knowledge captured perfectly
- 💬 Support - Users become advocates
- 📊 Analyst - Data drives decisions
- 📈 Marketer - Growth on autopilot
- 🎖️ Coordinator - Mission commander (PARALLEL STRIKE)
🚀 Quick Start (5 Minutes)
Installation
# Navigate to your project directory
cd /path/to/your/project
# Deploy AGENT-11
curl -sSL https://raw.githubusercontent.com/TheWayWithin/agent-11/main/project/deployment/scripts/install.sh | bash -s full
# Restart Claude Code
/exit
claude
Prerequisites: Project needs git repo, README, or package file (package.json, requirements.txt)
Verification
/agents # List deployed agents (should show 11)
@strategist What should we build first? # Test an agent
First Mission
From BOS-AI docs: /foundations init → /architect → /bootstrap → /coord continue (full workflow)
Existing project: /coord dev-alignment (analyzes codebase, creates docs)
New project: /coord dev-setup vision.md (initializes project structure)
Build feature: /coord build requirements.md
→ Complete setup guide: Essential Setup Guide for installation troubleshooting, MCP setup, and detailed verification steps.
🎯 Common Workflows
Learn the most common AGENT-11 workflows with real examples, time estimates, and expected outcomes.
Quick Reference
| Workflow | Command | Duration | Use Case |
|---|---|---|---|
| MVP | /coord mvp vision.md |
1-3 days | Rapid prototyping with auth, payments, deployment |
| Bug Fix | /coord fix bug-report.md |
2-4 hours | Root cause analysis, fix, regression tests |
| Design Review | /design-review |
1-2 hours | 7-phase UI/UX audit (RECON Protocol) |
| Feature | /coord build requirements.md |
4-8 hours | Complete feature with tests and docs |
| Security | /coord security |
2-3 hours | Vulnerability scanning, compliance checks |
| Refactor | /coord refactor target.md |
2-4 hours | Code quality improvement, tech debt |
| Performance | /coord optimize metrics.md |
3-6 hours | Bottleneck identification, optimization |
| Deploy | /coord deploy |
1-2 hours | Production deployment with monitoring |
Workflow Details
1. MVP Development - Build production-ready MVP in 1-3 days
- Full stack: auth, payments, testing, deployment
- Example: LLM.txt Mastery built in 3 days
- Cost: $5-10 in API usage
2. Bug Fixes - Debug and resolve production issues in 2-4 hours
- Root cause analysis, fix implementation, regression tests
- Cost: $0.50-1.50 in API usage
3. Design Review - Comprehensive UI/UX audit in 1-2 hours
- 7-phase RECON Protocol: Interaction, responsive, visual, accessibility, robustness, performance
- Deliverables: Issues classified by severity, screenshots, compliance report
- Cost: $0.50-1 in API usage
4. Feature Development - Build complete features in 4-8 hours
- Strategist → Architect → Developer → Tester → Documenter workflow
- Deliverables: Code, tests, API docs, deployment checklist
- Cost: $1-3 in API usage
5. Security Audit - Comprehensive security review in 2-3 hours
- OWASP Top 10, dependency scanning, compliance validation (SOC2, GDPR)
- Deliverables: Vulnerability report, fixes, security documentation
- Cost: $2-4 in API usage
6. Code Refactoring - Improve code quality in 2-4 hours
- Technical debt reduction, performance improvements
- All tests pass (no functionality changes)
- Cost: $0.75-2 in API usage
7. Performance Optimization - Tune application performance in 3-6 hours
- Bottleneck identification, query optimization, caching strategy
- Deliverables: Before/after benchmarks, performance report
- Cost: $1.50-3 in API usage
8. Production Deployment - Safe deployment in 1-2 hours
- Pre-deployment tests, zero-downtime deploy, monitoring, rollback plan
- Deliverables: Live deployment, smoke tests, runbook
- Cost: $0.50-1 in API usage
→ Complete workflow guide: Common Workflows Guide for:
- Detailed mission breakdowns with agent sequences
- Real-world examples and expected outcomes
- Recovery strategies for interrupted missions
- Verification procedures and deliverable checklists
More Workflows
AGENT-11 includes 20 pre-built missions covering every development need:
Setup & Initialization:
- DEV-SETUP - New project initialization (30-45 min)
- DEV-ALIGNMENT - Understand existing project (45-60 min)
- CONNECT-MCP - Setup external integrations (45-90 min)
Development Operations:
- REFACTOR - Code quality improvement (2-4 hours)
- DEPLOY - Production deployment (1-2 hours)
- DOCUMENT - Documentation creation (2-4 hours)
- MIGRATE - System migration (4-8 hours)
Performance & Quality:
- OPTIMIZE - Performance tuning (3-6 hours)
- INTEGRATE - Third-party APIs (3-6 hours)
- RELEASE - Release management (2-4 hours)
→ See all 20 missions with commands
→ Mission execution guide - Complete execution manual
⚙️ Essential Setup
Beyond basic installation, configure advanced features for production readiness.
Testing Infrastructure
Automatic setup during missions. Manual: @tester "Set up testing infrastructure"
MCP Integration
Running out of context during long missions? AGENT-11's MCP system optimizes context usage for your workflow.
🚀 Dynamic Tool Loading (v5.2.0+ - Recommended)
The new dynamic MCP tool loading system eliminates manual profile switching with 93% context reduction:
| Approach | Initial Context | Reduction |
|---|---|---|
| All Tools (static) | 51,000 tokens | baseline |
| Profile-based (legacy) | 3,000-15,000 tokens | 40-80% |
| Dynamic Loading | 3,300 tokens | 93% |
How it works: Agents discover and load tools on-demand using Tool Search. No profile switching required.
Quick Setup (Dynamic):
# Use dynamic MCP configuration
cp project/mcp/dynamic-mcp.json .mcp.json
# Configure API keys
cp .env.mcp.template .env.mcp
# Edit .env.mcp with your credentials
# Restart Claude Code
Documentation:
- 📖 MCP Setup Guide - Complete setup and usage instructions
- 🔄 Migration Guide - Migrate from profiles to dynamic loading
- 🔧 Troubleshooting - Common issues and solutions
Legacy: Profile-Based System
Profile-based MCP switching is still supported. See MCP-PROFILES.md for documentation.
Project Initialization
New: /coord dev-setup ideation.md (30-45 min) | Existing: /coord dev-alignment (45-60 min)
🏗️ How AGENT-11 Works
Understanding the architecture helps you maximize effectiveness through visual diagrams and clear explanations.
System Architecture Overview
graph TB
subgraph User["👤 User Layer"]
U[Developer/Founder]
end
subgraph Mission["🎯 Mission Layer"]
CMD["/coord Command"]
M1["20 Pre-built Missions"]
M2["Custom Workflows"]
end
subgraph Coord["🎖️ Coordination Layer"]
C[Coordinator Agent]
PP["project-plan.md<br/>(Planning)"]
AC["agent-context.md<br/>(Knowledge)"]
HN["handoff-notes.md<br/>(Handoffs)"]
PR["progress.md<br/>(History)"]
end
subgraph Agents["🤖 Specialist Layer"]
S[Strategist]
A[Architect]
D[Developer]
T[Tester]
O[Operator]
Des[Designer]
Doc[Documenter]
Sup[Support]
Ana[Analyst]
Mar[Marketer]
end
subgraph Tools["🔧 Tool Layer"]
R[Read/Write/Edit]
B[Bash/Git]
MCP["MCP Integrations<br/>(GitHub, Playwright, etc.)"]
Mem["Memory API<br/>(/memories/)"]
end
U --> CMD
CMD --> M1
CMD --> M2
M1 --> C
M2 --> C
C --> PP
C --> AC
C --> HN
C --> PR
C --> S
C --> A
C --> D
C --> T
C --> O
C --> Des
C --> Doc
C --> Sup
C --> Ana
C --> Mar
S --> R
A --> R
D --> R
D --> B
T --> R
O --> B
O --> MCP
Des --> MCP
Doc --> R
S --> Mem
A --> Mem
D --> Mem
style U fill:#e1f5ff
style CMD fill:#fff4e6
style C fill:#f3e5f5
style S fill:#e8f5e9
style A fill:#e8f5e9
style D fill:#e8f5e9
style T fill:#e8f5e9
style O fill:#e8f5e9
style Des fill:#e8f5e9
style Doc fill:#e8f5e9
style Sup fill:#e8f5e9
style Ana fill:#e8f5e9
style Mar fill:#e8f5e9
style Mem fill:#fffde7
Three-Layer Architecture Explained:
- Mission Layer - 20 pre-built missions + custom workflows via
/coord - Coordination Layer - Context preservation, parallel execution, dependency resolution
- Specialist Layer - 11 agents with domain expertise, tool permissions, extended thinking, self-verification
Agent Collaboration Flow
How specialists work together with zero context loss:
sequenceDiagram
participant U as 👤 User
participant C as 🎖️ Coordinator
participant S as 🎯 Strategist
participant A as 🏗️ Architect
participant D as 💻 Developer
participant T as ✅ Tester
participant AC as 📝 agent-context.md
participant HN as 🤝 handoff-notes.md
U->>C: /coord build requirements.md
C->>AC: Read mission context
C->>HN: Initialize handoff system
Note over C: Phase 1: Analysis
C->>S: Analyze requirements
S->>AC: Read existing context
S->>S: Create user stories (30-45 min)
S->>HN: Write findings for Architect
S->>AC: Update with decisions
Note over C: Phase 2: Design
C->>A: Design solution
A->>AC: Read strategist findings
A->>HN: Read specific handoff
A->>A: Create architecture (30-45 min)
A->>HN: Write specs for Developer
A->>AC: Update with design decisions
Note over C: Phase 3: Implementation
C->>D: Implement feature
D->>AC: Read all prior context
D->>HN: Read architect specs
D->>D: Write code + tests (2-4 hours)
D->>HN: Document implementation
D->>AC: Update with code decisions
Note over C: Phase 4: Validation
C->>T: Validate quality
T->>AC: Read full mission context
T->>HN: Read developer notes
T->>T: Run comprehensive tests (1 hour)
T->>HN: Report test results
T->>AC: Final quality assessment
C->>U: Mission complete ✅
Note over AC,HN: Zero context loss!<br/>100% knowledge preserved
Key Collaboration Principles:
- Sequential Handoffs - Each agent reads context from previous agents via handoff-notes.md
- Cumulative Knowledge - All decisions stored in agent-context.md for mission-wide awareness
- Zero Loss - Every finding, decision, and issue preserved across the entire mission
Context Management System
How AGENT-11 preserves 100% of knowledge across sessions and agent handoffs:
flowchart TD
subgraph Input["📥 Mission Input"]
MI["/coord build requirements.md"]
end
subgraph Context["🧠 Context Files"]
AC["agent-context.md<br/>━━━━━━━━━━━<br/>• Mission objectives<br/>• All findings (cumulative)<br/>• Technical decisions<br/>• Known issues<br/>• Dependencies"]
HN["handoff-notes.md<br/>━━━━━━━━━━━<br/>• Immediate task<br/>• Critical context<br/>• Warnings/blockers<br/>• Specific instructions<br/>• Test results"]
ER["evidence-repository.md<br/>━━━━━━━━━━━<br/>• Screenshots<br/>• Code snippets<br/>• Test results<br/>• API responses<br/>• Error logs"]
end
subgraph Memory["💾 Persistent Memory"]
MEM["/memories/<br/>━━━━━━━━━━━<br/>project/<br/> - overview.xml<br/> - architecture.xml<br/>technical/<br/> - patterns.xml<br/> - decisions.xml<br/>lessons/<br/> - insights.xml"]
end
subgraph Agents["🤖 Agents"]
A1[Agent reads context]
A2[Agent performs work]
A3[Agent updates context]
end
subgraph Output["📤 Deliverables"]
PP["project-plan.md<br/>(What we're doing)"]
PR["progress.md<br/>(What we did)"]
CODE["Code, Tests, Docs"]
end
MI --> AC
MI --> HN
AC --> A1
HN --> A1
MEM --> A1
A1 --> A2
A2 --> A3
A3 --> AC
A3 --> HN
A3 --> ER
A3 --> MEM
AC --> PP
AC --> PR
HN --> PP
HN --> PR
A2 --> CODE
style AC fill:#e3f2fd
style HN fill:#f3e5f5
style ER fill:#fff3e0
style MEM fill:#e8f5e9
style A1 fill:#fffde7
style A2 fill:#fffde7
style A3 fill:#fffde7
Context Preservation Protocol:
- Before Task - Agent reads agent-context.md + handoff-notes.md + /memories/
- During Task - Agent maintains awareness of all prior decisions
- After Task - Agent updates handoff-notes.md with findings for next specialist
- Coordinator - Merges findings into agent-context.md for mission-wide knowledge
Result: Zero context loss, 100% knowledge retention across sessions
→ Complete context management guide
Mission Execution Lifecycle
From command to deliverables - the complete mission workflow:
stateDiagram-v2
[*] --> Planning: /coord build requirements.md
Planning --> Analysis: Create project-plan.md
note right of Planning
Coordinator reads mission
Creates execution plan
Initializes context files
end note
Analysis --> Design: Strategist completes
note right of Analysis
Requirements analysis
User story creation
Success criteria defined
30-45 minutes
end note
Design --> Implementation: Architect completes
note right of Design
System design
Technology selection
Architecture documentation
30-45 minutes
end note
Implementation --> Testing: Developer completes
note right of Implementation
Code implementation
Unit test creation
Documentation updates
2-4 hours
end note
Testing --> Verification: Tests pass
note right of Testing
E2E test execution
Quality validation
Regression checks
1 hour
end note
Verification --> Complete: All verified
note right of Verification
Self-verification protocols
Deliverable confirmation
Progress documentation
15 minutes
end note
Testing --> Implementation: Tests fail
Implementation --> Implementation: Bugs found
Complete --> [*]: Mission success ✅
note right of Complete
progress.md updated
Learnings stored in /memories/
Ready for next mission
end note
Mission Phases Explained:
- Planning (5-10 min) - Coordinator reads mission, creates project-plan.md with phases
- Analysis (30-45 min) - Strategist analyzes requirements, creates user stories
- Design (30-45 min) - Architect designs system, selects technologies
- Implementation (2-4 hours) - Developer writes code, creates tests
- Testing (1 hour) - Tester validates quality, runs comprehensive test suite
- Verification (15 min) - Self-verification protocols ensure completeness
Total Time: 4-8 hours for standard feature development
Key Concepts
Context Preservation: Zero loss via agent-context.md + handoff-notes.md + /memories/. Extended Thinking: Ultrathink/Think harder/Think modes (39% effectiveness improvement). Tool Permissions: Least-privilege security model (64% read-only agents). Parallel Execution: Independent tasks run simultaneously for speed.
→ Memory Management | → Extended Thinking | → Tool Permissions
BOS-AI Integration
Workflow: BOS-AI (strategy) → AGENT-11 (execution)
# 1. Copy BOS-AI outputs to ideation/
# 2. Initialize: /coord dev-setup ideation/PRD.md
# 3. Build: /coord build ideation/PRD.md
→ Complete BOS-AI Integration Guide
🚀 Features & Capabilities
Complete overview of AGENT-11's capabilities organized by category.
Context Management
100% knowledge retention with native memory API, zero-loss context preservation (agent-context.md, handoff-notes.md), and strategic /clear usage (84% token reduction enabling 30+ hour missions).
→ Memory Management | → Context Editing
Project Management
Dual-file tracking (project-plan.md, progress.md with complete fix history), 20 pre-built mission workflows via /coord, automatic greenfield/brownfield bootstrap with CLAUDE.md generation.
→ Progress Tracking | → Bootstrap Guide | → Mission Library
Quality Assurance
SENTINEL Mode with separation of duties (zero bugs reach production), self-verification protocols (50% rework reduction), comprehensive testing (mcp__playwright, cross-browser, WCAG AA+).
→ Testing Guide | → Enhanced Prompting
Advanced Capabilities
Extended thinking modes (Ultrathink/Think harder/Think hard/Think - 39% effectiveness improvement), 15+ MCP integrations (GitHub, Playwright, Context7, Firecrawl, Supabase), RECON Protocol design reviews with 7-phase audit, OpsDev workflow with staging validation.
→ Extended Thinking | → Tool Permissions | → UI Doctrine
Performance Metrics
| Capability | Measurement | Impact |
|---|---|---|
| Mission Success | +15% (v4.0) | Opus 4.6 coordinator intelligence |
| Cost Efficiency | -24% total cost | Opus iterations offset token cost |
| Agent Effectiveness | 39% improvement | Extended thinking + self-verification |
| Token Efficiency | 84% reduction | Context editing + memory optimization |
| MCP Context | 60-90% reduction | Lean profile system (v4.1) |
| Autonomous Operation | 30+ hours | Multi-day missions without intervention |
| Rework Reduction | 50% fewer errors | Pre-handoff verification catches issues |
| Context Retention | 100% persistence | Zero knowledge loss across sessions |
| Time to MVP | 2-4 weeks | vs 3-6 months traditional |
| Cost per Feature | <$500 | vs $10-50k traditional |
→ Complete Performance Analysis
Known Limitations
AGENT-11 is powerful but not unlimited. Understanding these constraints helps you work effectively within the system.
1. Large Codebases
- Issue: Projects with >50 files may exceed context window during analysis
- Workaround: Break work into modules, use phased approach with multiple missions
- Command:
@coordinator "Focus only on [specific module/feature]"
2. Complex Dependencies
- Issue: Some packages require manual configuration (native bindings, platform-specific tools)
- Workaround: Let agents set up what they can, complete manual steps with provided instructions
- Example: Database setup, API key configuration, external service accounts
3. Mission Interruptions
- Issue: If mission fails mid-way, context may be partially lost
- Workaround: Check progress.md for completed work, resume from last checkpoint
- Command:
@coordinator "Resume from last checkpoint in progress.md"
4. API Rate Limits
- Issue: Claude API rate limits may pause long-running missions (>2 hours)
- Workaround: Break into smaller phases, use
/clearbetween phases to reset context - Recovery: Wait 1-5 minutes, then continue with next phase
5. Context Size Management
- Issue: Very large files (>2000 lines) or long conversations exceed context window
- Workaround: Use Glob/Grep for targeted searches instead of reading entire files
- Command:
@coordinator "Search for [specific pattern] instead of reading whole file"
6. Browser Automation Limitations
- Issue: Some JavaScript-heavy sites may not work with Playwright automation
- Workaround: Use manual testing for complex UI, automated for standard paths
- When: Design reviews of SPAs, PWAs, or sites with heavy client-side rendering
7. Platform-Specific Issues
- Issue: Some commands work differently on Windows vs Mac vs Linux
- Workaround: Use cross-platform alternatives (npm scripts vs shell scripts)
- Example: Prefer
npxover direct binary calls, Node.js scripts over bash
8. Agent Specialization Boundaries
- Issue: Some tasks don't fit cleanly into one agent's role
- Workaround: Use
/coordmissions for cross-functional work requiring multiple agents - Example: Security audit needs @architect + @developer + @tester coordination
9. Learning Curve
- Issue: First-time users need 1-2 hours to understand agent system and workflows
- Workaround: Start simple - use
/meetingfor conversations, progress to single @agent commands, then to/coordmissions - Progression: Individual agents → simple missions → complex orchestration
10. Single-User Operation
- Issue: No real-time collaboration between multiple humans on same project
- Workaround: Use git branches for parallel work, merge through standard PR process
- Best Practice: One person runs missions, team reviews results via PR
11. Internet Dependency
- Issue: Requires active internet connection for Claude API access
- Workaround: Work offline on manual tasks, run missions when connection available
- Impact: No offline development assistance
12. Token Costs
- Issue: API usage costs vary by mission complexity ($0.50-$10 per mission)
- Workaround: Start with smaller missions to estimate costs, use
/coordefficiently - Monitoring: Check API usage in Claude dashboard, set budget alerts
→ Complete capabilities and limitations guide
Feature Documentation
3,750+ lines of comprehensive guides: Memory Management | Bootstrap | Context Editing | Extended Thinking | Tool Permissions | Enhanced Prompting
🆘 Getting Unstuck Protocol
Quick Recovery: /clear + @coordinator "I'm stuck" | System Check: /agents, ls .claude/agents/ | Test: @developer "Create hello world HTML" | Escalation: Troubleshooting | @support | GitHub Issues | Discord
📚 Complete Documentation
🚀 Quick Start
📖 Setup & Configuration
- Project Deployment | Update Installation | MCP Integration | MCP Troubleshooting | Greenfield | Brownfield
🎯 Mission Execution
🏗️ Architecture & Planning
🆕 Modernization Guides (v2.0-v4.1)
- Memory Management | Bootstrap Guide | Context Editing | Extended Thinking | Tool Permissions | Enhanced Prompting
- v4.0+: Model Selection Guide | MCP Optimization Guide
🎨 Quality & Design
📚 Agent & Mission Reference
🛠️ Templates
💬 Getting Help
@support| Troubleshooting | Success Stories | GitHub Issues | Discord
📊 Mission Progress Tracking System
AGENT-11 uses a dual-file tracking system: project-plan.md (forward-looking roadmap) and progress.md (backward-looking changelog with complete issue history including all fix attempts). This approach enables 10-40x faster resolution of similar future issues.
→ Complete Progress Tracking Guide
🔄 Project Lifecycle Management
Three-tier cleanup strategy prevents context pollution: Milestone Transition (every 2-4 weeks), Project Completion (when shipping), or Continue Active Work. Lessons extracted to searchable repository while keeping active context lean (200 lines vs 20,000+).
→ Complete Lifecycle Management Guide | → Cleanup Checklist
🧪 Testing & Quality Assurance
Separation of duties: Tester analyzes (read-only), Developer implements tests, Tester validates. SENTINEL Mode runs 7-phase protocol for critical releases. Zero bugs reach production with 80%+ coverage on critical paths.
🚀 Your First Mission
Pattern: /coord [mission] [input-file.md] or @agent for direct access. Examples: /coord build requirements.md | /coord fix bug-report.md | /coord mvp vision.md
→ Mission Execution Guide | → Mission Library
🎮 Command Reference
AGENT-11 provides 12 slash commands for the complete development workflow:
📋 Planning Commands (BOS-AI → Execution)
These commands transform foundation documents into executable project plans:
📥 /foundations - Document Processing
Extract structured data from BOS-AI foundation documents into agent-consumable YAML
/foundations init # Initial extraction of all documents
/foundations refresh # Sync changes (new, modified, removed docs)
/foundations status # Show document states and checksums
/foundations validate # Verify completeness before building
Subcommands:
| Command | When to Use | What It Does |
|---|---|---|
init |
First time setup | Scans documents/foundations/, extracts ALL to structured YAML |
refresh |
After any document changes | Syncs directory → detects new, modified, removed documents |
status |
Check current state | Shows all documents with checksums and modification status |
validate |
Before /bootstrap |
Verifies required docs present and extractions complete |
Foundation → YAML Mapping (9 categories):
| Foundation Document | Output YAML | Schema Contains |
|---|---|---|
| prd.md, -prd-.md, requirements.md | prd.yaml |
Features, tech stack, metrics, business rules, state machines |
| vision-mission.md, vision.md | vision.yaml |
Vision, mission, hedgehog concept, core values |
| strategic-roadmap.md, roadmap.md | roadmap.yaml |
Phases, deliverables, milestones, success criteria |
| client-success-blueprint.md, icp.md | icp.yaml |
Personas, pain points, JTBD, anti-personas |
| market-and-client-research.md, research.md | research.yaml |
Competitive analysis, market research |
| brand-style-guidelines.md, brand.md | brand.yaml |
Colors, typography, components, shadows, animations |
| positioning-statement.md, positioning.md | positioning.yaml |
Core positioning, competitive differentiation, messaging |
| marketing-bible.md, go-to-market.md | marketing.yaml* |
Go-to-market, channels, launch strategy |
| pricing-strategy.md, pricing.md | pricing.yaml |
Tiers, Marketing Physics, value ladder, upgrade triggers |
*marketing.yaml only generated if dedicated marketing document exists (not positioning)
Common Workflows:
# Initial setup (first time)
cp ~/BOS-AI-output/*.md documents/foundations/
/foundations init
# After editing documents
/foundations refresh # Only re-extracts what changed
# After adding new document (e.g., pricing)
cp ~/BOS-AI-output/pricing-strategy.md documents/foundations/
/foundations refresh # Detects and extracts new document
# Before building
/foundations validate # Ensures everything is ready
Output: .context/structured/*.yaml files + handoff-manifest.yaml
🏛️ /architect - System Design
Generate architecture.md with all technical decisions before planning
/architect # Interactive mode selection (recommended)
/architect --mode engaged # Walk through 7 decision areas
/architect --mode auto # Use PRD defaults
/architect --stack nextjs-supabase # Use predefined stack profile
When to Use Each Mode:
| Mode | Best For | Time | Recommended When |
|---|---|---|---|
| Engaged | New projects, first-time users | 10-15 min | You want to understand trade-offs and make informed decisions |
| Auto | Regeneration, experienced users | 1-2 min | Your PRD is detailed and you trust the defaults |
Mode Selection (when you run /architect without flags):
How would you like to proceed?
1. Auto Mode - Use PRD defaults
2. Engaged Mode - Walk through decisions (recommended)
Engaged Mode - What You'll Decide (7 areas):
| Decision | What You Choose | Example Options |
|---|---|---|
| 1. Application Architecture | Overall structure | Monolith (MVP), Modular Monolith, Microservices |
| 2. Frontend Stack | Framework + styling | Next.js 14 + Tailwind + shadcn/ui |
| 3. Backend & Database | Database + API approach | Supabase + RLS, multi-tenancy strategy |
| 4. Authentication | Auth provider + methods | Clerk/Supabase Auth, OAuth providers, roles |
| 5. External Integrations | Services + billing model | Stripe subscription, AI models, email (Resend) |
| 6. Infrastructure | Hosting + CI/CD | Vercel, Railway, GitHub Actions |
| 7. Security & Observability | Monitoring + security | Sentry, PostHog, rate limiting |
What it generates:
- Comprehensive
architecture.md(~400 lines) - System diagrams and component relationships
- Decision log with rationale for each choice
- Integration patterns and security policies
Prerequisites: Run /foundations init first
Example Engaged Mode Question:
🏗️ Decision 1/7: Application Architecture
Your PRD indicates a web application with these features:
- User authentication
- Dashboard with data visualization
- AI-powered features
Options:
1. Monolith (Recommended for MVP)
Fast to build, easy to deploy, refactor later
2. Modular Monolith
Clear boundaries, easier to split later
3. Microservices
⚠️ Overkill for MVP - adds operational complexity
Select [1/2/3]:
🏗️ /bootstrap - Plan Generation
Transform foundation documents + architecture.md into executable project-plan.md
/bootstrap # Interactive mode selection (recommended)
/bootstrap --mode engaged # Review PRD assumptions with checkpoints
/bootstrap --mode auto # Generate immediately (no questions)
/bootstrap --mode preview # Preview without writing files
/bootstrap --mode auto --type api # Auto mode with explicit project type
When to Use Each Mode:
| Mode | Best For | Time | Recommended When |
|---|---|---|---|
| Engaged | New projects, first-time users | 3-5 min | You want to validate PRD extraction and priorities |
| Auto | Regeneration, trusted PRDs | 30 sec | Your PRD is validated and you want speed |
| Preview | Inspection | 30 sec | You want to see the plan before committing |
Mode Selection (when you run /bootstrap without flags):
How would you like to proceed?
1. Auto Mode - Generate immediately (fast, no questions)
2. Engaged Mode - Review assumptions first (recommended for new users)
3. Preview Mode - Show what would be generated
Engaged Mode - What You'll Review (5 checkpoints):
| Checkpoint | What You Verify | Why It Matters |
|---|---|---|
| 1. Foundation Summary | Product name, description, P0 features | Catches extraction errors early |
| 2. Tech Stack | Framework versions, integrations | Ensures correct dependencies |
| 3. Priority Validation | P0/P1/P2 feature assignments | Controls MVP scope |
| 4. Phase Structure | Phase breakdown, task count | Validates effort estimates |
| 5. Final Confirmation | All decisions summary | Last chance before generation |
What it generates:
- Phased
project-plan.mdwith rolling wave detail - Phase 1: Fully detailed with tasks, acceptance criteria, and quality gates
- Phase 2+: Outlined with key milestones (detailed when you reach them)
- Quality gates configured per project type (blocking/warning/info)
Prerequisites: Run /foundations init and /architect first
Example Engaged Mode Checkpoint:
📋 Checkpoint 2/5: Tech Stack Confirmation
From your PRD and architecture.md:
Frontend: Next.js 14 (App Router)
Styling: Tailwind CSS + shadcn/ui
Backend: Supabase (PostgreSQL + RLS)
Auth: Clerk
Payments: Stripe
Is this correct? [Y/n/edit]:
📊 /plan - Project State Management
View and manage project state, phases, and quality gates
/plan status # Current phase, progress, blockers
/plan next # What's coming up
/plan phase 2 # Deep dive into Phase 2
/plan gate 2 # Quality gate status for Phase 2
/plan update task 3 complete # Mark task complete
/plan archive # Archive completed phases
Subcommands:
| Command | Purpose |
|---|---|
status |
Progress overview, blockers, next action |
next |
Upcoming tasks and phase previews |
phase N |
Detailed phase breakdown |
gate N |
Quality gate validation status |
update |
Update task status, add blockers |
archive |
Archive completed phases for context savings |
🧰 /skills - SaaS Skill Discovery
Discover and manage production-ready SaaS code patterns
/skills # List all available skills
/skills saas-auth # Show auth skill details
/skills match "stripe payments" # Find matching skills
/skills stack # Show current stack profile
Available Skills (auto-loaded by Coordinator based on task keywords):
| Skill | Triggers | Tokens |
|---|---|---|
saas-auth |
auth, login, oauth | ~3,800 |
saas-payments |
stripe, checkout, subscription | ~4,200 |
saas-multitenancy |
tenant, org, rls | ~4,100 |
saas-billing |
billing, plan, quota | ~3,900 |
saas-email |
email, resend, notification | ~3,200 |
saas-onboarding |
onboarding, wizard | ~3,500 |
saas-analytics |
analytics, posthog | ~3,600 |
How it works: When task descriptions contain trigger keywords, the Coordinator automatically loads the relevant skill and delegates to the appropriate specialist with production-ready patterns.
🚀 Execution Commands
🎖️ /coord - Mission Orchestration
Orchestrate multi-agent missions with automatic specialist coordination
# Pattern: /coord [mission] [input-file.md]
/coord build requirements.md # Feature development (4-8 hours)
/coord fix bug-report.md # Bug resolution (1-3 hours)
/coord mvp vision.md # Build MVP (1-3 days)
/coord dev-setup ideation.md # Project initialization (30-45 min)
/coord dev-alignment # Analyze existing project (45-60 min)
What it does:
- Automatically delegates to appropriate specialists
- Maintains context across agent handoffs
- Updates project-plan.md and progress.md
- Ensures quality through built-in verification
🤝 /meeting - Strategic Conversations
Have natural, conversational discussions with specialists for brainstorming and planning
# Pattern: /meeting @agent [topic]
/meeting @strategist # Product strategy discussion
/meeting @architect "microservices vs monolith" # Architecture consultation
/meeting @designer "improving user onboarding" # Design brainstorming
/meeting @developer "real-time data synchronization" # Technical problem-solving
When to use:
- Brainstorming new features or approaches
- Getting expert advice before formal missions
- Exploring technical decisions and trade-offs
- Creative collaboration and idea generation
Available specialists: @strategist, @architect, @coordinator, @developer, @tester, @designer, @operator, @marketer, @analyst, @documenter, @support
🎨 /design-review - UI/UX Audit
Comprehensive design audit using RECON Protocol with automated testing
/design-review # Full 7-phase UI/UX audit with evidence
What it analyzes (7-phase RECON Protocol):
- Preparation - Environment setup, baseline screenshots
- Interaction - User flows, micro-interactions, navigation
- Responsive - Mobile, tablet, desktop compatibility
- Visual - Typography, spacing, hierarchy, consistency
- Accessibility - WCAG AA+ compliance, screen readers
- Robustness - Edge cases, error states, loading
- Performance - Load times, console errors
Deliverables:
- Issues classified by severity (BLOCKER/HIGH/MEDIUM/NITPICK)
- Screenshots and reproduction steps
- Accessibility compliance report
- Specific recommendations with examples
🔍 /recon - Design Intelligence
Quick design reconnaissance for understanding existing UI/UX patterns
/recon # Rapid design analysis and pattern identification
What it does:
- Analyzes current design system and patterns
- Identifies UI/UX strengths and weaknesses
- Documents design decisions and rationale
- Provides strategic design recommendations
Use when: Need quick design assessment before major changes or starting new features
📊 /report - Progress Reports
Generate stakeholder-ready progress reports automatically
# Pattern: /report [since_date]
/report # Last 7 days (default)
/report 2025-08-20 # Since specific date
Report includes:
- Executive summary of progress
- Completed tasks with business impact
- Issues encountered and resolutions
- Current project status and health
- Key metrics and trends
- Next milestones and resource needs
- BOS-AI alignment (if integrated)
Output: Creates progress-report.md with professional formatting suitable for stakeholders, clients, or management
🔬 /pmd - Post Mortem Analysis
Systematic root cause analysis of failures to prevent recurrence
# Pattern: /pmd [issue_description]
/pmd # Analyze recent failures from progress.md
/pmd "Coordinator not using Task tool" # Analyze specific issue
/pmd "Installation failing on Windows" # Analyze deployment failure
Analysis covers:
- Agent Performance - Prompt clarity, scope compliance, tool usage
- Documentation Quality - CLAUDE.md accuracy, task definitions
- Tool Usage - MCP prioritization, error handling
- Process Issues - Planning, communication, testing coverage
Deliverables:
- Timeline of events leading to failure
- Root cause analysis with evidence
- Immediate fixes (do now)
- Short-term improvements (this week)
- Long-term enhancements (this month)
- Prevention strategies and detection mechanisms
Output: Creates post-mortem-analysis.md with actionable recommendations
📅 /dailyreport - Daily Progress Reports + Social Media Posts
Turn today's progress.md entries into a structured daily report plus voice-aligned blog, Twitter/X, and LinkedIn posts for build-in-public sharing
/dailyreport # Extract today's work from progress.md and generate all four files
Report includes:
- Completed milestones grouped by context (features, fixes, infrastructure, docs)
- Issues & learnings with root cause analysis and failed-attempt history
- Lessons learned and patterns noticed
- Next steps for tomorrow
Output Files (4 files per day):
progress/
├── 2026-04-11.md # Raw daily report (source of truth)
├── 2026-04-11-blog.md # Long-form blog post (voice-aligned)
├── 2026-04-11-twitter.md # Twitter/X post (copy-paste ready)
└── 2026-04-11-linkedin.md # LinkedIn post (copy-paste ready)
Social Media Features:
- Twitter/X: 180-260 char sweet spot, dual-link OG preview structure, 1-2 hashtags
- LinkedIn: 800-1000 char sweet spot, hook-optimized first 140 chars, genuine closing question
- Copy-paste ready: character count validated,
{{PRODUCT_URL}}placeholder for final edit - Voice-aligned: every post follows your voice guide (see below)
Claude-native architecture: /dailyreport is pure Claude Code — no Python script,
no OPENAI_API_KEY, no external API costs. Claude reads your progress.md, drafts the
structured report, and derives all three derived outputs following the voice guide to
the letter. This is the same pipeline /blog uses.
Use when:
- End of work day for daily build-in-public updates
- After major milestone completion
- When switching between multiple projects
- For authentic progress transparency
Benefits:
- Blog-ready markdown format
- Platform-optimized social posts (Twitter/X + LinkedIn)
- Shows both wins and challenges, including failed fix attempts
- Documents learnings for future reference
- No API keys, no usage costs, works immediately after install
Voice Alignment: /dailyreport and /blog share a voice guide system. Drop avoice-guide.md in your project root (or set DAILYREPORT_VOICE_GUIDE=/path/to/file.md)
and every generated post follows your voice. The built-in default is tuned for authentic,
non-AI-sounding build-in-public writing. See the voice alignment docs
for the full resolution chain.
✍️ /blog - Topic-Driven Blog Posts + Social Media Posts
Draft a long-form blog post and matching Twitter/X + LinkedIn versions on any topic, pulling context from your repo
# Inline topic
/blog why I stopped using feature flags for small teams
# Brief file
/blog briefs/cron-incident.md
# Topic with explicit context files
/blog the file persistence bug --context progress/2025-11-19.md post-mortem-analysis.md
/dailyreport turns a structured progress log into a build-in-public update. /blog
is for everything else: opinion pieces, tutorials, deep dives, post-mortems, reactions
— any writing that isn't tied to a daily changelog. Both commands share the same
Claude-native pipeline, the same voice guide file, and the same social post format.
The only difference is the shape of the input: /dailyreport parses structured
progress logs, /blog takes any topic you want to write about.
Output Files (3 files per post):
blog/
├── 2026-04-11-feature-flags-small-teams.md # Long-form blog post (400-1200 words)
├── 2026-04-11-feature-flags-small-teams-twitter.md # Twitter/X (180-260 char, copy-paste ready)
└── 2026-04-11-feature-flags-small-teams-linkedin.md # LinkedIn (800-1000 char, copy-paste ready)
What Claude does on each run:
- Resolves the voice guide (same chain as
/dailyreport— custom file, env var, or default) - Reads
README.md,CLAUDE.md, the brief file (if given), any--contextfiles, and
intelligently pulls in related files (e.g.progress.mdfor a bug story) - Drafts the long-form post following every voice rule
- Derives Twitter/X and LinkedIn versions from the long-form post
- Runs an AI-tell scrub and rewrites any banned vocabulary before writing files
- Writes all three files to
blog/and previews the Twitter post inline
When to use /blog vs /dailyreport:
| Situation | Command |
|---|---|
End-of-day progress update from progress.md |
/dailyreport |
| Opinion piece, tutorial, deep dive, post-mortem story | /blog |
| Build-in-public changelog narrative | /dailyreport |
| Reaction or explainer on any topic | /blog |
Both commands write in the same voice and share the same voice guide file. Define your
voice once, use it everywhere.
Shared configuration (same env vars as /dailyreport):
DAILYREPORT_VOICE_GUIDE=/path/to/voice-guide.md # Custom voice guide (optional)
DAILYREPORT_BASE_URL=yourdomain.com # Used in social post blog links
See full /blog documentation for argument forms, drafting
rules, troubleshooting, and examples.
📦 /planarchive - Intelligent Context Management
Automatically archive completed work from tracking files using semantic analysis and smart triggers
# Pattern: /planarchive [options]
/planarchive # Interactive mode with analysis
/planarchive --analyze # Show what would be archived with reasoning
/planarchive --aggressive # Archive ALL completed work
/planarchive --target-lines=1200 # Archive until reaching specific size
/planarchive --check-duplicates # Find duplicated content
What it analyzes (5 intelligent triggers):
- Completed Sprint Rule - Sprints marked COMPLETE + >7 days old
- Size-Based Rule - Sections >500 lines that are complete
- Dated Phase Rule - Phases with COMPLETE status + old dates
- Duplication Rule - Content duplicated in .env.example, architecture.md
- Historical Detail Rule - Code snippets/guides from completed work
Archival Scoring System:
Score = (age_days × 0.3) + (size_lines × 0.2) + (completion_status × 0.5)
- Score ≥0.8: High priority (archive immediately)
- Score 0.5-0.8: Medium priority
- Score <0.5: Low priority (keep active)
What gets archived:
- project-plan.md: Completed sprints, resolved risks, locked decisions
- progress.md: Entries older than threshold (default: 14 days), resolved issues
What stays active:
- Executive summary and current objectives
- Active work (pending tasks, unresolved issues)
- Last 7 entries minimum
- Content from last 7 days (unless --aggressive)
Key Features:
- Semantic Analysis - Understands completion status, dates, section hierarchy
- Auto-Summaries - Generates 2-3 sentence summaries for archived content
- Token Optimization - Reports estimated token savings
- Safety First - Automatic backups, reversible operations, validation checks
Token Budget Levels:
- ✅ Optimal: <15,000 tokens
- ⚠️ Warning: 25,000-35,000 tokens (should archive)
- 🚨 Critical: >35,000 tokens (mandatory archival)
Command Options:
--analyze- Show archival analysis with scores and reasoning (dry run)--aggressive- Archive ALL completed work regardless of age--target-lines=N- Archive until reaching specific line count--check-duplicates- Identify content duplicated across files--days=N- Archive entries older than N days (default: 14)--no-summaries- Keep full content instead of generating summaries
Output:
📊 Archive Analysis Complete
project-plan.md:
- Sprint 1: 692 lines, age 25d, score 0.92 → ARCHIVE
- Sprint 2: 487 lines, age 18d, score 0.85 → ARCHIVE
Total: 1,179 lines (61% reduction)
progress.md:
- 28 entries >14 days (890 lines) → ARCHIVE
Total: 890 lines (47% reduction)
Estimated token savings: ~4,250 tokens (63% reduction)
Archive Locations:
project-plan-archive.md- Archived phases, sprints, completed tasksprogress-archive.md- Archived progress entries, resolved issues
Use when:
- End of sprint/phase (natural archive point)
- Before major mission (start with clean context)
- Files exceed 25,000 tokens (context overhead)
- Monthly maintenance (regular hygiene)
Benefits:
- Keeps tracking files lean and efficient (<1,200 lines ideal)
- Reduces token overhead for faster agent reads
- Preserves complete history in archive files
- Zero data loss - all content preserved
Command Comparison
Planning Commands (BOS-AI → Project Plan):
| Command | Purpose | Duration | Output | Best For |
|---|---|---|---|---|
/foundations |
Process & sync BOS-AI documents (init/refresh/status/validate) | 2-5 min | Structured YAML | Starting from BOS-AI |
/architect |
Design system architecture (Auto/Engaged modes) | 10-15 min | architecture.md | Technical decisions before building |
/bootstrap |
Generate project plan (Auto/Engaged/Preview modes) | 1-5 min | project-plan.md | Plan-driven development |
/plan |
View/manage project state | Instant | Status, gates, progress | Tracking & management |
/skills |
Discover SaaS patterns | Instant | Skill details | Finding code patterns |
Execution Commands (Building & Shipping):
| Command | Purpose | Duration | Output | Best For |
|---|---|---|---|---|
/coord |
Execute missions with specialists | 1-8 hours | Code, tests, docs | Building, fixing, deploying |
/meeting |
Strategic conversations | 15-60 min | Discussion, ideas | Brainstorming, planning |
/design-review |
Full UI/UX audit | 1-2 hours | Design report | Pre-launch quality checks |
/recon |
Quick design analysis | 30-45 min | Design insights | Understanding existing UI |
/report |
Stakeholder updates | 5-10 min | Progress report | Communication, tracking |
/pmd |
Failure analysis | 30-60 min | Root cause analysis | Learning from mistakes |
/dailyreport |
Daily progress + social posts | 2-5 min | Blog + Twitter + LinkedIn | Build-in-public updates |
/blog |
Topic-driven blog + social posts | 2-5 min | Blog + Twitter + LinkedIn | Opinion pieces, tutorials, deep dives |
/planarchive |
Intelligent tracking file cleanup | 2-5 min | Archive files | Token optimization, context management |
🎖️ Community Success Stories
"Each project gets its own elite team that understands the codebase perfectly. Game changer."
— Alex Chen, Solo Founder ($15k MRR)
"Project-local agents mean no confusion, no context switching. They just get it."
— Sarah Martinez, Technical Founder
"Clean deployments, isolated teams per project. Finally, AI that scales with my workflow."
— David Kim, Multi-Project Founder
"The v2.0 memory system is incredible - agents remember decisions from days ago. No more context loss."
— AGENT-11 Beta Tester
🔥 Mission Library (20 Missions)
Setup Missions (NEW!)
- 🚀 DEV-SETUP - Greenfield project initialization (30-45 min)
- 🎯 DEV-ALIGNMENT - Existing project understanding (45-60 min)
- ⚙️ OPSDEV-SETUP - DevOps & environment configuration (20-30 min)
- 📋 CLAUDE-SETUP - CLAUDE.md creation & sync (15-25 min)
- 🔌 CONNECT-MCP - MCP discovery and connection (45-90 min)
Development Missions
- 🏗️ BUILD - New feature development (4-8 hours)
- 🐛 FIX - Emergency bug resolution (1-3 hours)
- ♻️ REFACTOR - Code quality improvement (2-4 hours)
- 🚀 DEPLOY - Production deployment (1-2 hours)
- 📚 DOCUMENT - Documentation creation (2-4 hours)
- 📐 ARCHITECTURE - System architecture documentation (2-3 hours)
Strategic Missions
- 📋 PRODUCT-DESCRIPTION - Product definition with risk management (2-3 hours)
- 💡 MVP - Minimum viable product (1-3 days)
- 🔄 MIGRATE - System migration (4-8 hours)
- ⚡ OPTIMIZE - Performance optimization (3-6 hours)
- 🔒 SECURITY - Security audit & fixes (4-6 hours)
- 🔌 INTEGRATE - Third-party integration (3-6 hours)
- 🎯 RELEASE - Release management (2-4 hours)
Mission Command Quick Reference
| Mission | Command | Input Required | Duration | Use Case |
|---|---|---|---|---|
| Setup Missions | ||||
| DEV-SETUP | /coord dev-setup ideation.md |
✅ Ideation | 30-45 min | New project initialization |
| DEV-ALIGNMENT | /coord dev-alignment |
❌ None | 45-60 min | Understand existing project |
| CONNECT-MCP | /coord connect-mcp |
❌ None | 45-90 min | Setup external integrations |
| Core Development | ||||
| BUILD | /coord build requirements.md |
✅ Requirements | 3-8 hours | Feature development |
| MVP | /coord mvp vision.md |
✅ Vision | 6-12 hours | Minimum viable product |
| FIX | /coord fix bug-report.md |
✅ Bug Report | 1-4 hours | Bug resolution |
| REFACTOR | /coord refactor |
❌ None | 2-4 hours | Code improvement |
| DEPLOY | /coord deploy |
❌ None | 1-2 hours | Production deployment |
| DOCUMENT | /coord document |
❌ None | 2-4 hours | Documentation creation |
| MIGRATE | /coord migrate requirements.md |
✅ Requirements | 4-8 hours | System migration |
| Specialized Operations | ||||
| ARCHITECTURE | /coord architecture vision.md |
✅ Vision | 2-4 hours | System design |
| OPTIMIZE | /coord optimize |
❌ None | 3-6 hours | Performance tuning |
| SECURITY | /coord security |
❌ None | 4-6 hours | Security audit |
| INTEGRATE | /coord integrate requirements.md |
✅ Requirements | 3-6 hours | Third-party APIs |
| RELEASE | /coord release |
❌ None | 2-4 hours | Deployment prep |
| Business & Growth | ||||
| PRODUCT-DESCRIPTION | /coord product-description vision.md |
✅ Vision | 2-3 hours | Marketing copy |
| Operations | ||||
| GENESIS | /coord genesis |
❌ None | 1-2 hours | Project reconnaissance |
| RECON | /coord recon |
❌ None | 1-3 hours | Design intelligence |
Input File Templates: Available in /templates/mission-inputs/
Legend: ✅ Required input file | ❌ No input needed
Mission MCP Profile Guide
Some missions benefit from specific MCP profiles for optimal performance and context efficiency:
- Testing missions (BUILD with E2E tests, QA phases) →
/mcp-switch testing(adds Playwright) - Deployment missions (DEPLOY, RELEASE) →
/mcp-switch deployment(Netlify + Railway access) - Database work (MIGRATE, schema changes) →
/mcp-switch database-stagingor/mcp-switch database-production - Full-stack development (complex features) →
/mcp-switch fullstack(all development MCPs) - General development (BUILD, FIX, REFACTOR) →
/mcp-switch core(lightweight, 80% less context)
Quick Commands:
- Switch profiles:
/mcp-switch [profile-name] - See all profiles:
/mcp-list - Check current:
/mcp-status
🔄 Updating Existing Installation
Already have AGENT-11? Get the latest features including the /coord mission system:
# Single command to update your project
cd /path/to/your/project
curl -sSL https://raw.githubusercontent.com/TheWayWithin/agent-11/main/project/deployment/scripts/install.sh | bash -s core
🛡️ Your CLAUDE.md is Safe! AGENT-11 uses a two-file architecture:
Your Project
├── CLAUDE.md ← Your personal preferences (NEVER touched)
└── .claude/
└── CLAUDE.md ← AGENT-11 library instructions (managed by installer)
- ✅ Your root
/CLAUDE.mdis never modified or overwritten - ✅ AGENT-11 installs to
.claude/CLAUDE.mdonly - ✅ Existing
.claude/CLAUDE.mdis backed up before updating - ✅ Both files are read automatically by Claude Code
📋 CLAUDE.md Architecture Guide → - Full details on the two-file system
🛠️ Troubleshooting
Complete Troubleshooting Guide →
💖 Support AGENT-11
AGENT-11 is a passion project that's 100% free and always will be. If this framework has:
- ⚡ Saved you weeks of development time
- 🚀 Helped you ship your MVP faster
- 💰 Eliminated the need for expensive contractors
Consider supporting its development:
Your support keeps AGENT-11 free, actively maintained, and growing with new capabilities!
Other ways to support:
- ⭐ Star the repo to help others discover it
- 📢 Share AGENT-11 with fellow founders and developers
- 🐛 Report bugs and suggest features
- 📝 Contribute to documentation
🤝 Join the Elite
AGENT-11 is open source and community-driven. We welcome contributions from fellow operators.
📜 License
MIT - Use AGENT-11 to build your empire.
Ready to deploy your project squad?
🚀 Project Setup · 📖 Quick Start · 📚 Full Documentation
"One project. One founder. Eleven specialists. Unlimited potential."
Yorumlar (0)
Yorum birakmak icin giris yap.
Yorum birakSonuc bulunamadi