An agent skill behaves predictably to the exact degree its boundary is specified. Treat it as a capability list and it drifts; treat it as a contract (declared inputs, outputs, and scope), and it behaves like a well-designed API.
把 token 當設計變數而非月底帳單:粗放的任務成本沒有上限,但分太細也不會更省。快取讓過度切分反而更貴,重點是找到對的顆粒度。
Governance has a bounded, knowable token cost; ungoverned agent work tends not to. And task granularity has its own price: caching can make over-decomposition cost more than it looks.
No evidence, no completion: the one rule that closes most AI agent failures. A task isn't done until it produces a verifiable artifact (commit SHA, test output).
AI 代理每個新對話都失憶？Work Log 用一份 markdown 記錄任務進度與決策，讓 Claude Code 跨 session 接續，不用每次重講背景。
AI agent governance maps onto distributed systems patterns: audit logs, delivery acknowledgment, idempotency, least privilege. The prior art already exists.
不用框架也能治理 AI 代理：靠 AGENTS.md / CLAUDE.md 記憶檔、evidence 習慣和範圍宣告，就能擋掉大部分 Claude Code、Cursor 的常見問題。
Why AI agents fail: most agent failures aren't model problems, they're governance problems, and the two need completely different fixes. How to tell them apart.
AI 代理（AI Agent）開發常見問題整理：輸出難核查、跳步驟、跨對話失憶、範圍失控。從實戰痛點到 Agentic OS 的應對方向，附 Claude Code 實例。
Most teams treat AI skills as saved prompt templates. They're not. A proper Skill is a capability contract—with input schemas, tool bindings, validation, and retry logic. Here's what that distinction means for production system design.