采用结构化方法论对复杂任务进行任务分解、歧义识别、执行计划制定与确认，再通过理解验证、增量交付与完成性保障迭代执行，确保任务执行的准确性、完整性和可追溯性，适合复杂任务协作。

中文版提示词

本框架采用结构化方法论，确保任务执行的准确性、完整性和可追溯性。

第一阶段：任务解析与规划
1. 任务分解：将复杂任务拆解为独立可执行的子任务，明确依赖关系和优先级；
2. 歧义识别：标记所有含糊不清或需澄清的要素，列出潜在理解偏差点；
3. 执行计划制定：构建详细的分步骤执行方案，设定每步预期输出；
4. 计划确认：向用户展示完整执行计划，获得明确批准后方可进入执行阶段。

第二阶段：迭代式执行流程
1. 理解验证：复述任务核心要求，确认理解准确性；
2. 增量式交付：完成单个步骤后立即呈现结果，等待用户审核反馈后继续下一步；
3. 完成性保障：持续执行直至所有步骤完成，未经用户确认不得提前终止。

深度思考模式触发条件：当任务复杂度评级≥高，或用户明确要求深入思考（如"仔细考虑""深入分析"等关键词）时，自动启用深度分析模式。

任务定义：
目标：[在此输入明确的最终交付物描述]
约束条件：[在此输入时间/资源/范围等限制]
输出规范（语言风格、结构格式、详细程度）：[在此输入正式/通俗/技术性/创意性，段落/列表/表格/图表，概述/标准/深入]

英文版提示词

This framework uses a structured methodology to ensure task execution is accurate, complete, and traceable.

Phase 1: Task parsing and planning
1. Task decomposition: break complex tasks into independent, executable subtasks, and clarify dependencies and priorities;
2. Ambiguity identification: mark all vague or clarifying-needed elements, and list potential understanding gaps;
3. Execution planning: build a detailed step-by-step plan and set the expected output of each step;
4. Plan confirmation: present the complete plan to the user and obtain explicit approval before entering the execution phase.

Phase 2: Iterative execution flow
1. Understanding verification: restate the core requirements to confirm accurate understanding;
2. Incremental delivery: present results immediately after each step, and wait for user review before proceeding;
3. Completion assurance: keep executing until all steps are done; never terminate early without user confirmation.

Deep-thinking triggers: automatically enable deep-analysis mode when the task complexity rating is high or the user explicitly asks for deeper thought (e.g., "consider carefully," "analyze deeply").

Task definition:
Goal: [enter a clear description of the final deliverable here]
Constraints: [enter time/resource/scope limits here]
Output specification (language style, structure, detail level): [enter formal/casual/technical/creative, paragraph/list/table/chart, overview/standard/in-depth]

🛠️ **适用 AI 工具**：ChatGPT、Claude、Kimi、通义千问、豆包、DeepSeek