{
  "id": "page:docs-reference-repos-wshobson-agents-research",
  "_kind": "Page",
  "_file": "wiki/docs/reference-repos/wshobson/agents/research.md",
  "_cluster": "wiki",
  "attributes": {
    "nodeKind": "Page",
    "sourcePath": "docs/reference-repos/wshobson/agents/research.md",
    "sourceKind": "repo-docs",
    "title": "wshobson/agents",
    "displayName": "wshobson/agents",
    "slug": "docs/reference-repos/wshobson/agents/research",
    "articlePath": "wiki/docs/reference-repos/wshobson/agents/research.md",
    "article": "\n# wshobson/agents\n\n**Repository**: https://github.com/wshobson/agents  \n**Stars**: 33,494 ⭐  \n**License**: MIT  \n**Language**: Python  \n**Archetype**: Claude Code Multi-Agent Orchestration Framework  \n**Processing Date**: 2026-04-13\n\n## Overview\n\nA comprehensive production-ready system combining 182 specialized AI agents, 16 multi-agent workflow orchestrators, 149 agent skills, and 96 commands organized into 77 focused, single-purpose plugins for Claude Code. This represents one of the largest organized collections of agent orchestration patterns and workflows.\n\n## Repository Structure\n\n```\nwshobson/agents/\n├── .claude-plugin/          # Claude Code plugin configuration\n├── plugins/                 # 77 focused plugins (main content)\n│   ├── accessibility-compliance/\n│   ├── agent-orchestration/\n│   └── ...\n├── docs/                    # Comprehensive documentation\n│   ├── plugins.md           # Plugin reference\n│   ├── agents.md            # Agent reference\n│   ├── agent-skills.md      # Skills documentation\n│   ├── usage.md             # Usage patterns\n│   └── architecture.md      # Design principles\n└── tools/                   # Supporting tooling\n```\n\n## Key Features & Architectural Patterns\n\n### Granular Plugin Architecture\n- **77 focused plugins** optimized for minimal token usage and composability\n- Average 3.6 components per plugin (follows Anthropic's 2-8 pattern)\n- Each plugin completely isolated with its own agents, commands, and skills\n- Install only what you need - no unnecessary resources loaded into context\n\n### Agent Orchestration Models\n- **182 specialized agents** across domain expertise areas\n- **16 workflow orchestrators** for complex multi-agent operations\n- **Progressive disclosure** - skills load knowledge only when activated\n- **Three-tier model strategy** for Opus 4.6, Sonnet 4.6 & Haiku 4.5\n\n### Skills System (149 Skills)\n- Modular knowledge packages with progressive disclosure\n- Token efficiency through selective loading\n- Specialized expertise across domains\n- Compatible with agent orchestration patterns\n\n## Notable Orchestration Patterns\n\n### Multi-Agent Review Systems\n- `comprehensive-review` plugin with architect-review, code-reviewer, security-auditor\n- Parallel analysis with different perspectives\n- Coordinated feedback synthesis\n\n### Full-Stack Orchestration\n- `full-stack-orchestration` plugin for complex development workflows\n- Multi-agent coordination for end-to-end development\n- Phase-based execution with handoffs\n\n### Agent Teams (Experimental)\n- Parallel code review with specialized reviewers\n- Hypothesis-driven debugging with multiple agents\n- Research teams with parallel investigation streams\n\n### Context-Driven Development (Conductor Plugin)\n- Structured Context → Spec & Plan → Implement workflow\n- Track-based development with semantic organization\n- TDD workflow with verification checkpoints\n\n## Quality Framework (PluginEval)\n\nThree-layer evaluation framework:\n- **Static analysis** (instant)\n- **LLM judge** (semantic)\n- **Monte Carlo simulation** (statistical)\n\nQuality dimensions:\n- Triggering accuracy, orchestration fitness, output quality\n- Scope calibration, progressive disclosure, token efficiency\n- Robustness, structural completeness, ecosystem coherence\n\n## Technology Integration\n\n- **Claude Code** as primary harness\n- **Agent Skills** for knowledge management\n- **Plugin marketplace** distribution model\n- **Makefile** for development automation\n- **CI/CD** integration patterns\n\n## Extractable Processes\n\n### Core Orchestration Patterns\n1. **Multi-perspective review processes** - parallel expert analysis\n2. **Progressive skill disclosure** - token-efficient knowledge loading\n3. **Plugin composition strategies** - modular system assembly\n4. **Agent handoff protocols** - structured workflow transitions\n\n### Development Methodologies\n1. **Context-driven development** - vision → spec → implementation\n2. **Track-based organization** - semantic work unit management\n3. **Verification checkpoint patterns** - quality gates in workflows\n4. **Semantic revert strategies** - logical undo operations\n\n### Quality Assurance Frameworks\n1. **Multi-layer evaluation** - static, semantic, statistical analysis\n2. **Anti-pattern detection** - automated quality assessment\n3. **Statistical rigor** - Wilson score CI, bootstrap validation\n4. **CI gate integration** - threshold-based quality control\n\n## Value for Babysitter Marketplace\n\n**High Value** - This repository contains the most comprehensive collection of agent orchestration patterns discovered, with well-documented architectural principles and proven workflows. The progressive disclosure patterns, multi-agent coordination strategies, and quality evaluation frameworks provide substantial value for babysitter process development.\n\n### Key Extractable Elements\n- Multi-agent orchestration patterns for complex workflows\n- Progressive disclosure mechanisms for token efficiency\n- Quality evaluation frameworks with statistical rigor\n- Plugin composition and modular architecture principles\n\n## Classification\n\n**Category**: Multi-Agent Orchestration Framework  \n**Subcategory**: Claude Code Plugin Ecosystem  \n**Complexity**: Enterprise-grade  \n**Maturity**: Production-ready  \n**Adoption**: High (33K+ stars, active development)\n\n## Research Notes\n\nThis repository represents the current state-of-the-art in Claude Code agent orchestration, with sophisticated approaches to:\n- Token efficiency through modular loading\n- Multi-agent coordination patterns\n- Quality measurement and validation\n- Enterprise-scale deployment patterns\n\nThe architectural principles and orchestration patterns here could significantly inform babysitter's multi-agent workflow capabilities.\n\n## Library Mapping\n\n| Extractable Process | Library Status | Action | Existing Path | Target Placement |\n|-------------------|----------------|--------|---------------|------------------|\n| Multi-Perspective Review Processes | NEW | Parallel expert analysis with perspective coordination | - | specializations/shared/multi-perspective-review.js |\n| Progressive Skill Disclosure | NEW | Token-efficient knowledge loading | - | specializations/shared/progressive-disclosure.js |\n| Agent Handoff Protocols | NEW | Structured workflow transitions between specialized agents | - | specializations/shared/agent-handoff-protocols.js |\n| Context-Driven Development | NEW | Vision → spec → implementation methodology | - | methodologies/context-driven-development/ |\n| Track-Based Organization | NEW | Semantic work unit management | - | specializations/shared/track-based-organization.js |\n| Multi-Layer Quality Evaluation | NEW | Static, semantic, statistical analysis framework | - | specializations/shared/multi-layer-evaluation.js |\n| Plugin Composition Strategies | NEW | Modular system assembly patterns | - | specializations/shared/plugin-composition.js |\n| Agent Team Coordination | NEW | Multi-agent parallel coordination patterns | - | specializations/shared/agent-team-coordination.js |\n\n## Plugin Marketplace Mapping\n\n| Plugin Idea | Marketplace Status | Action | Existing Plugin | Target Placement |\n|-------------|-------------------|--------|-----------------|------------------|\n| Progressive Disclosure Engine | NEW | Token-efficient knowledge loading system | - | plugins/a5c/marketplace/blueprints/progressive-disclosure-engine/ |\n| Context-Driven Development | NEW | Vision-to-implementation workflow automation | - | plugins/a5c/marketplace/blueprints/context-driven-development/ |\n| Multi-Layer Quality Framework | NEW | Comprehensive quality evaluation with statistical rigor | - | plugins/a5c/marketplace/blueprints/multi-layer-quality-framework/ |\n",
    "documents": []
  },
  "outgoingEdges": [],
  "incomingEdges": [
    {
      "from": "page:docs-reference-repos",
      "to": "page:docs-reference-repos-wshobson-agents-research",
      "kind": "contains_page"
    }
  ]
}