Type: url
Original: https://waylandz.com/ai-agent-book/%E7%AC%AC29%E7%AB%A0-Agentic-Coding/
Status: generated
Language: zh
Summary
A chapter of an online book on AI agent architecture that explains agentic coding as agents that understand a whole codebase, plan changes, edit code, run tests and iterate, in contrast to code completion. It covers codebase understanding (keyword plus semantic search, dependency graphs), anchor-based editing, sandboxed execution, the edit-test-fix loop, version-control integration, routing tasks across model tiers, security controls and common pitfalls, using the Shannon agent platform’s source code as a running example.