Agentic Harness Patterns（模式与技能）

The model loop is easy. User -> LLM -> tool_use -> execute -> loop fits on a napkin. What makes a production agent actually work — reliably, safely, at scale — is everything around the loop: memory that persists across sessions, permissions that fail closed, context budgets that don't explode, multi-agent…

课时

• Claude Code 源码蒸馏 - Harness Engineering 实践记录
• 六层 Harness 结构
  • 记忆系统
• 深度参考文档
  • memory-persistence-pattern
  • skill-runtime-pattern
  • tool-registry-pattern
  • permission-gate-pattern
  • agent-orchestration-pattern
  • context-engineering
  • select-pattern
  • compress-pattern
  • isolate-pattern
  • hook-lifecycle-pattern
  • task-decomposition-pattern
  • bootstrap-sequence-pattern
• Agentic Harness Patterns（模式与技能）
• 文档
  • Distilling Claude Code Source — A Harness Engineering Practice Log
• skills
  • agentic-harness-patterns
    • Agentic Harness Patterns
    • references
      • Agent Orchestration Pattern
      • Bootstrap Sequence Pattern
      • Context Engineering Pattern
      • Hook Lifecycle Pattern
      • Memory and Persistence Pattern
      • Permission Gate Pattern
      • Skill Runtime and Packaging Pattern
      • Long-running Work Management
      • Tool Registry Pattern
      • context-engineering
        • Context Compression and Snapshot Management
        • Context Isolation for Delegated Work
        • Context Selection and Progressive Disclosure

开始学习 → Claude Code 源码蒸馏 - Harness Engineering 实践记录