{
  "id": "page:docs-reference-repos-shareai-lab-learn-claude-code-research",
  "_kind": "Page",
  "_file": "wiki/docs/reference-repos/shareai-lab/learn-claude-code/research.md",
  "_cluster": "wiki",
  "attributes": {
    "nodeKind": "Page",
    "sourcePath": "docs/reference-repos/shareAI-lab/learn-claude-code/research.md",
    "sourceKind": "repo-docs",
    "title": "shareAI-lab/learn-claude-code",
    "displayName": "shareAI-lab/learn-claude-code",
    "slug": "docs/reference-repos/shareai-lab/learn-claude-code/research",
    "articlePath": "wiki/docs/reference-repos/shareAI-lab/learn-claude-code/research.md",
    "article": "\n# shareAI-lab/learn-claude-code\n\n- **Archetype**: harness-framework (educational harness architecture)\n- **Stars**: 52,165\n- **Last pushed**: 2026-04-07\n- **License**: MIT\n- **Discovered**: 2026-04-12\n- **Source**: gh-search (topic: claude-code)\n- **Skills found**: 4 (pdf, code-review, mcp-builder, agent-builder)\n\n## Summary\nA \"nano claude code-like agent harness\" built from scratch in Python/Bash for educational purposes. Includes skills for PDF processing, code review, MCP server building, and agent building. The agent-builder skill contains a comprehensive agent design methodology: core philosophy (\"the model already knows how to be an agent\"), three elements (capabilities, knowledge, context), and design thinking framework.\n\n## Assessment\nHIGH VALUE. The agent-builder skill is a standout -- it contains a well-articulated agent design methodology that emphasizes simplicity and the model's inherent capability. The code-review skill and MCP-builder skill contain procedural workflows. The educational framing provides clear, extractable process definitions.\n\n## Extraction Priority\nHIGH -- The agent-builder methodology is directly extractable:\n- Agent design thinking -> methodologies/ (agent architecture methodology)\n- Code review skill -> already have code-review but may contain unique patterns\n- MCP-builder -> specializations/shared/ (cross-domain MCP development process)\n\n## Processes\n1. **agent-design-thinking** -- Purpose -> Domain -> Capabilities (3-5 essential) -> Knowledge -> Context analysis\n2. **agent-build-loop** -- The core agent loop: context + capabilities -> decide -> act or respond\n3. **mcp-server-development** -- Process for building MCP servers from scratch\n4. **code-review-workflow** -- Structured code review process\n\n## Plugin Ideas\nNone specific -- methodology content, not tool integration.\n\n## Harness Integration Ideas\n\n### Educational Architecture Patterns\n- **Current limitation**: Limited documentation of harness architecture principles\n- **Integration approach**: Use educational patterns to improve babysitter harness documentation and architecture understanding\n- **Implementation scope**: Internal documentation, architecture refinement, team education\n\n### Agent Loop Optimization\n- **Current limitation**: Could benefit from clearer agent loop architecture documentation\n- **Integration approach**: Apply educational agent loop patterns to babysitter runtime documentation\n- **Implementation scope**: `packages/babysitter-sdk/src/runtime/`, improved loop architecture\n\n### Permission System Enhancement\n- **Current limitation**: Limited safety validation patterns for model intent execution\n- **Integration approach**: Study permission layer patterns for agent execution safety\n- **Implementation scope**: `packages/babysitter-sdk/src/security/`, model intent validation\n\n## Implicit Procedural Knowledge\n- \"Start with 3-5 capabilities, add more only when agent consistently fails\" -- capability minimalism principle\n- \"Make knowledge available, not mandatory. Load when relevant, not upfront\" -- lazy knowledge loading\n- \"Context is precious. Isolate noisy subtasks. Truncate verbose outputs. Protect clarity\" -- context hygiene\n- The agent loop distilled to its simplest form\n- \"The model does the reasoning. The harness gives the model a working environment\" -- fundamental harness philosophy\n\n## Library Mapping\n\n| Extractable Process | Library Status | Action | Existing Path | Target Placement |\n|-------------------|----------------|--------|---------------|------------------|\n| Agent Design Thinking Methodology | NEW | Purpose → Domain → Capabilities → Knowledge → Context analysis methodology | - | methodologies/agent-design-thinking/ |\n| Agent Build Loop Process | NEW | Core agent loop: context + capabilities → decide → act or respond | - | specializations/shared/agent-build-loop.js |\n| MCP Server Development Process | NEW | Process for building MCP servers from scratch with structured workflow | - | specializations/shared/mcp-server-development.js |\n| Code Review Workflow | EXISTING | Structured code review process | specializations/shared/code-review.js | specializations/shared/code-review-workflow.js |\n| Capability Minimalism Process | NEW | Start with 3-5 capabilities, add more only when agent consistently fails | - | specializations/shared/capability-minimalism.js |\n| Lazy Knowledge Loading | NEW | Make knowledge available, not mandatory; load when relevant, not upfront | - | specializations/shared/lazy-knowledge-loading.js |\n| Context Hygiene Process | NEW | Context management: isolate noisy subtasks, truncate verbose outputs, protect clarity | - | specializations/shared/context-hygiene.js |\n\n## Plugin Marketplace Mapping\n\n| Plugin Idea | Marketplace Status | Action | Existing Plugin | Target Placement |\n|-------------|-------------------|--------|-----------------|------------------|\n| Agent Design Assistant | NEW | Educational agent design methodology with capability minimalism and context hygiene principles | - | plugins/a5c/marketplace/blueprints/agent-design-assistant/ |\n| MCP Development Suite | NEW | MCP server development toolkit with structured workflow and educational patterns | - | plugins/a5c/marketplace/blueprints/mcp-development-suite/ |\n| Educational Harness Framework | NEW | Simplified harness architecture patterns for learning and documentation | - | plugins/a5c/marketplace/blueprints/educational-harness-framework/ |\n",
    "documents": []
  },
  "outgoingEdges": [],
  "incomingEdges": [
    {
      "from": "page:docs-reference-repos",
      "to": "page:docs-reference-repos-shareai-lab-learn-claude-code-research",
      "kind": "contains_page"
    }
  ]
}