Learn AI-assisted development through progressive levels. Start with 50-line commands that just work, advance to full automation when ready.
# Clone to your plugins directory
git clone https://github.com/boshu2/agentops.git ~/.claude/plugins/agentops
# Or add as marketplace
/plugin marketplace add boshu2/agentopsStart at Level 1. Each level adds ONE concept:
| Level | What You Learn | Commands |
|---|---|---|
| L1-basics | Run commands, see results | /research, /implement |
| L2-persistence | Save findings to .agents/ |
+ /retro |
| L3-state-management | Track work with issues | + /plan, beads |
| L4-parallelization | Execute waves in parallel | + /implement-wave |
| L5-orchestration | Full automation | + /autopilot |
# Your first command (L1)
/research "how does authentication work"
# Progress when ready
/implement # Make changes based on researchEach level has demo transcripts showing real sessions. See levels/L1-basics/demo/.
/research → /plan → /implement → /retro
↓ ↓ ↓ ↓
explore decompose execute learn
| Level | Workflow |
|---|---|
| L1 | /research → /implement (single session) |
| L2 | /research → /implement → /retro (with persistence) |
| L3 | /research → /plan → /implement <id> → /retro (with issues) |
| L4 | /research → /plan → /implement-wave → /retro (parallel) |
| L5 | /autopilot <epic> (hands-off) |
| Category | Count | Description |
|---|---|---|
| Levels | 5 | Progressive learning L1-L5 |
| Reference | 3 | PDC, FAAFO, Failure Patterns |
| Skills | 12 | Domain knowledge (55 areas consolidated) |
| Commands | 28 | Full reference versions |
| Profiles | 3 | Role-based configurations |
Deep framework content, consulted when needed:
| Document | Content |
|---|---|
| PDC Framework | Prevent, Detect, Correct methodology |
| FAAFO Alignment | Fast, Ambitious, Autonomous, Fun, Optionality |
| Failure Patterns | 12 ways AI-assisted development goes wrong |
Skills load into main context with full tool access—no sub-agent limitations.
| Skill | Triggers | Areas |
|---|---|---|
| languages | Python, Go, Rust, Java, TypeScript | 6 |
| development | API, backend, frontend, mobile, LLM | 8 |
| documentation | docs, README, OpenAPI, Diátaxis | 4 |
| code-quality | review, test, coverage | 3 |
| research | explore, find, analyze | 6 |
| validation | validate, verify, tracer bullet | 4 |
| operations | incident, debug, postmortem | 4 |
| monitoring | metrics, alerts, SLO | 2 |
| security | pentest, SSL, secrets | 2 |
| data | ETL, Spark, ML, MLOps | 4 |
| meta | context, session, workflow | 6 |
| specialized | accessibility, UX, risk | 6 |
Based on Vibe Coding by Gene Kim & Steve Yegge.
| Level | Trust | Use For |
|---|---|---|
| L5 | 95% | Formatting, linting |
| L4 | 80% | Boilerplate, config |
| L3 | 60% | Standard features |
| L2 | 40% | New features |
| L1 | 20% | Architecture, security |
| L0 | 0% | Novel exploration |
- Below 40% context → 98% success rate
- Above 60% context → 24% success rate
Fast (10-16x) · Ambitious (solo feasible) · Autonomous (team output) · Fun (50% flow) · Optionality (120x options)
agentops/
├── levels/ # 🎯 START HERE - Progressive learning
│ ├── L1-basics/ # Single-session, no state
│ ├── L2-persistence/ # Add .agents/ output
│ ├── L3-state-management/ # Add issue tracking
│ ├── L4-parallelization/ # Add wave execution
│ └── L5-orchestration/ # Full autopilot
├── reference/ # Framework documentation
│ ├── pdc-framework.md
│ ├── faafo-alignment.md
│ └── failure-patterns.md
├── .agents/ # AI memory system
│ ├── research/ # Deep exploration docs
│ ├── plans/ # Implementation roadmaps
│ ├── patterns/ # Reusable solutions
│ ├── learnings/ # Session insights
│ └── retros/ # Retrospectives
├── commands/ # Full reference commands (28)
├── skills/ # Domain knowledge (12)
├── profiles/ # Role configurations (3)
└── .beads/ # Git-based issue tracking
- Vibe Coding Book - Gene Kim & Steve Yegge
- Vibe Ecosystem - Full documentation
- vibe-check - Metrics CLI tool
- Kubernetes the Hard Way - Inspiration for progressive levels
MIT