A bilingual guide page for how AI agents plan, use tools, and collaborate in real work.
This page is a fast entry point for AI agent practice: scoped roles, task framing, tool use, delegation, review loops, and handoff patterns. Skim the guide layer here first, then open the PDF for the complete document.
Effective agent work starts with a clean operating frame: define the goal, bound the scope, give the agent the right tools, and make the next action obvious enough to execute safely.
The document also focuses on execution discipline: let agents break down work, let tools touch the environment, and let reviews catch weak assumptions before they turn into mistakes.
Start here for the big picture. The goal is to give teams a compact shared language for planning, tool-grounded execution, delegation, and review in agent-based workflows.
Strong agents do better when the task is framed cleanly: the goal is explicit, the context is relevant, and the next action is concrete enough to execute without guessing.
Real agent workflows should not stay trapped in chat. Tools let agents inspect files, search systems, edit code, run checks, and verify assumptions against the environment.
The document treats review and handoff as first-class parts of agent work, so results stay inspectable, transferable, and repeatable instead of depending on one lucky run.
This guide reads best from task framing to tools to review loops. That sequence makes it easier to turn abstract agent ideas into reliable operating behavior.
The original PDF is embedded below. Use this page for a fast conceptual pass, then keep reading here or open the PDF separately for the full document.