Production-ready patterns for AI-assisted development. The Research → Plan → Implement workflow, session persistence, and 55 specialized agents.
# Clone to your plugins directory
git clone https://github.com/boshu2/agentops.git ~/.claude/plugins/knowledge-os
# Or add as marketplace
/plugin marketplace add boshu2/agentops| Category | Count | Description |
|---|---|---|
| Commands | 29 | RPI workflow, sessions, bundles, quality, documentation |
| Agents | 55 | Specialized agents across 8 domains |
| Skills | 7 | Domain knowledge packs |
graph LR
R["Research"] -->|bundle| P["Plan"]
P -->|spec| I["Implement"]
I -->|patterns| L["Learn"]
L -.->|next| R
/research # Deep exploration before planning
/plan # Create implementation plan
/implement # Execute approved plan with validation
/session-start # Initialize session with context
/session-end # Save progress and summarize
/bundle-save # Save context for multi-session work
/bundle-load # Resume from saved context| Domain | Count | Examples |
|---|---|---|
| Code Quality | 8 | code-reviewer, test-generator, validation-planner |
| Languages | 7 | golang-pro, python-pro, typescript-pro, rust-pro, java-pro |
| Architecture | 6 | backend-architect, spec-architect, fullstack-developer |
| DevOps | 5 | deployment-engineer, incident-responder, error-detective |
| Documentation | 5 | api-documenter, doc-explorer, documentation-create-docs |
| Data/ML | 4 | data-engineer, ml-engineer, ai-engineer, mlops-engineer |
| Security | 3 | penetration-tester, risk-assessor, continuous-validator |
| Meta | 5 | context-manager, task-decomposition-expert, prompt-engineer |
55 agents (click to expand)
- accessibility-specialist
- ai-engineer
- api-documenter
- archive-researcher
- assumption-validator
- autonomous-worker
- backend-architect
- change-executor
- code-explorer
- code-review-improve
- code-reviewer
- connection-agent
- context-manager
- continuous-validator
- customer-support
- data-engineer
- data-scientist
- deployment-engineer
- doc-explorer
- document-structure-analyzer
- documentation-create-docs
- documentation-diataxis-auditor
- documentation-optimize-docs
- error-detective
- frontend-developer
- fullstack-developer
- golang-pro
- history-explorer
- incident-responder
- incidents-postmortems
- incidents-response
- ios-developer
- java-pro
- meta-memory-manager
- meta-observer
- meta-retro-analyzer
- ml-engineer
- mlops-engineer
- mobile-developer
- monitoring-alerts-runbooks
- network-engineer
- penetration-tester
- performance-engineer
- prompt-engineer
- python-pro
- risk-assessor
- rust-pro
- shell-scripting-pro
- spec-architect
- task-decomposition-expert
- test-generator
- tracer-bullet-deployer
- typescript-pro
- ui-ux-designer
- validation-planner
| Skill | Purpose |
|---|---|
| base | Foundation patterns and templates |
| brand-guidelines | Consistent documentation styling |
| doc-curator | Documentation quality management |
| git-workflow | Git best practices and automation |
| skill-creator | Create new skills from patterns |
| test-gap-scanner | Identify missing test coverage |
| testing | Test patterns and frameworks |
Trust calibration for AI collaboration:
| Level | Trust | Verification | Use For |
|---|---|---|---|
| L5 | 95% | Final only | Formatting, linting |
| L4 | 80% | Spot check | Boilerplate, config |
| L3 | 60% | Key outputs | Standard features |
| L2 | 40% | Every change | New features |
| L1 | 20% | Every line | Architecture, security |
| L0 | 0% | Research only | Novel exploration |
Based on Vibe Coding by Gene Kim & Steve Yegge.
| Letter | Meaning | Result |
|---|---|---|
| F | Fast | 10-16x productivity gains |
| A | Ambitious | Projects become feasible solo |
| A | Autonomous | Team-scale output from individuals |
| F | Fun | 50% more flow state time |
| O | Optionality | 120x more exploration options |
| Loop | Timeframe | Focus |
|---|---|---|
| Inner | Seconds | Individual prompts and responses |
| Middle | Hours | Work sessions or features |
| Outer | Days-weeks | Architecture and direction |
Context utilization matters:
- Below 40% → 98% success rate
- Above 60% → 24% success rate
Higher context causes AI to forget instructions and contradict itself.
Watch for: Tests Passing Lie, Fix Spiral, Eldritch Horror, Silent Deletion, Confident Hallucination, and others documented in the book.
- Vibe Coding Book - Gene Kim & Steve Yegge's methodology
- Vibe Ecosystem - Full ecosystem documentation
- vibe-check - Measure your vibe coding metrics
- bodenfuller.com/workflow - Watch the workflow in action
MIT