什么是 自主功能规划器？

自主功能规划器是 Openclaw Skills 生态系统中的专用工具，旨在进行深度的自主系统架构设计。通过将单个前置用户命令作为基础，该技能可以独立推导系统域，并将其系统地分解为独特的、可实现的功能。它专为需要全面、顺序执行计划且无需频繁往复提示的开发者而设计。

该技能确保每个规划的功能在功能上是唯一的，并遵守锁定的项目范围，使其成为使用 Openclaw Skills 进行高层项目概念化的核心资产。它作为一个边界输出的自主规划器运行，这意味着它在建立的域内对任务进行排序，而不会发明范围之外的内容。

下载入口:https://github.com/openclaw/skills/tree/main/skills/vishnubedi3/autonomous-feature-planner

安装与下载

1. ClawHub CLI

从源直接安装技能的最快方式。

2. 手动安装

将技能文件夹复制到以下位置之一

优先级：工作区 > 本地 > 内置

3. 提示词安装

将此提示词复制到 OpenClaw 即可自动安装。

自主功能规划器 应用场景

• 从粗略的项目想法过渡到详细的实施路线图。
• 为新软件产品生成技术要求的顺序列表。
• 将基础系统架构扩展为独立、可管理的各个功能模块。
• 在使用 Openclaw Skills 开发的初始发现阶段创建标准化的功能规范。

1. 从对话历史中捕捉基础命令，以锁定系统域并建立基准意图。
2. 声明推导项目范围所需的所有最低限度必要假设，并在会话中锁定。
3. 进入边界输出规划循环，选择正好一个符合锁定域的下一个功能。
4. 定义功能范围，包括明确的包含和排除项，以防止范围蔓延。
5. 制定具有特定输入、输出和依赖关系的线性、有序实施计划。
6. 为该功能说明一个可验证的成功条件并终止当前迭代。
7. 重复此过程，直到发出停止命令或规划完成。

自主功能规划器 配置指南

要在您的环境中激活此功能，请确保已配置 Openclaw Skills 框架。您可以通过提供系统描述后跟触发关键字来启动规划器。

# 在系统描述后触发技能
use autonomous-feature-planner

# 其他触发方式
start autonomous planning
autonomous expansion

要随时终止规划过程，请使用以下命令：

stop planning

自主功能规划器 数据架构与分类体系

该技能维持严格的、仅限追加的数据结构，以确保整个规划生命周期的一致性。数据按照以下分类进行顺序迭代组织：

name: autonomous-feature-planner description: > Autonomously plans and specifies system features starting from the user’s most recent command, continuing without further user input until explicitly stopped. Use when the user explicitly invokes autonomous planning to extend a system from a prior command. Trigger keywords: use autonomous-feature-planner, start autonomous planning, autonomous expansion, continuous feature planning.

Activation Criteria

Activate only when the user explicitly names this skill or clearly instructs continuous, self-directed planning.

Activation requires:

• Exactly one immediately preceding user command
• That command must describe or imply a system, product, or process

Do not activate if:

• The previous command is conversational, evaluative, or meta
• The previous command is itself a stop instruction
• The user requests execution, deployment, or real-world action

Foundation Handling

• The last user command before activation is the foundation.
• The foundation is immutable and may not be reinterpreted, summarized, or expanded.
• The foundation establishes the system domain and intent baseline.

If the foundation cannot define a system domain, fail immediately.

Execution Model

This skill operates as a bounded-output autonomous planner. Autonomy applies to sequencing, not to scope invention.

Initialization

1. Capture the foundation command verbatim.
2. Derive one explicit system domain statement.
3. Declare all assumptions required to derive the domain.
4. Lock the domain and assumptions for the entire session.

If assumptions exceed minimal necessity, fail.

Planning Loop

Each iteration must perform exactly the following steps:

1. Select exactly one next feature that:
   • Directly fits within the locked system domain
   • Is functionally distinct from all prior features
2. Define the feature scope using explicit inclusions and exclusions.
3. Produce a linear, ordered implementation plan with no branches.
4. Specify:
   • Required inputs
   • Produced outputs
   • Dependencies on prior features
5. State one verifiable success condition.
6. Terminate the iteration.

Only one feature is permitted per iteration. No iteration may reference future, unplanned features.

Output Rules

• Each iteration must be labeled sequentially.
• Output must be strictly structured and utilitarian.
• No summaries, retrospectives, vision statements, or meta-commentary.
• No repetition, restatement, or revision of earlier iterations.

Ambiguity Handling

• All ambiguity must be resolved during initialization.
• Resolution must favor the narrowest viable interpretation.
• No new assumptions may be introduced after initialization.

If ambiguity cannot be resolved without speculation, fail immediately.

Consistency Enforcement

• All output is append-only.
• Previously planned features are immutable.
• If a contradiction is detected, halt immediately with failure.

Scope and Runaway Prevention

• Features must not generate sub-features.
• Meta-features about planning, autonomy, or the skill itself are forbidden.
• Each iteration must be finite and self-contained.
• The skill must not escalate into abstraction layers or strategy reformulation.

Constraints & Non-Goals

• No execution, simulation, or external state changes.
• No file creation or modification.
• No user interaction during operation.
• No external tools, memory, or hidden state.
• No goal invention outside the locked domain.

Failure Behavior

Immediately halt and output a single failure message if:

• The foundation cannot define a coherent system domain
• Minimal assumptions are insufficient
• Internal consistency cannot be preserved
• Planning would require execution or unverifiable facts

No additional output is permitted after failure.

Stop Condition

Immediately stop all planning when the user issues any command containing:

• "stop autonomous-feature-planner"
• "stop planning"
• "disable autonomous-feature-planner"

After a stop command:

• Output exactly one character: "."
• Output no other text, whitespace, or newlines.