← 返回 Skills 市场
286
总下载
0
收藏
0
当前安装
13
版本数
在 OpenClaw 中安装
/install pad-mode
功能描述
Turn messy requests into structured plans. PAD Mode (Plan → Act → Deliver) gives your AI agent project management superpowers — automatic task breakdown, liv...
安全使用建议
This skill is instruction-only and appears coherent with its purpose: it will create and update plan files (plans/YYYY-MM-DD-<slug>.md) using the included template. Before installing, verify the source (the homepage is unknown and README points to a user repo) and only clone from a repository you trust. Be aware that the skill conceptually supports background/sub-agent execution — confirm your OpenClaw agent's permissions and limits for running sub-agents or background tasks if you do not want autonomous long-running operations. Finally, check the plan files the agent writes to your workspace to ensure they don't contain unexpected data or links.
功能分析
Type: OpenClaw Skill
Name: pad-mode
Version: 1.2.0
The 'pad-mode' skill is a structured project management framework for OpenClaw agents, implementing a Plan-Act-Deliver (PAD) workflow. The skill's logic, defined in SKILL.md, focuses on increasing agent reliability through mandatory user approval gates, explicit execution mode selection (foreground/background), and detailed task verification. It uses a dedicated workspace directory (~/.openclaw/workspace/plans/) to maintain state and includes strict instructions to prevent the agent from proceeding without user consent, which acts as a safety guardrail rather than a malicious injection.
能力标签
能力评估
Purpose & Capability
The name/description (PAD Mode: Plan → Act → Deliver) aligns with the SKILL.md and included assets: it creates plan files from a template, tracks task status, and enforces approval gates. There are no unrelated environment variables, binaries, or config paths requested.
Instruction Scope
Runtime instructions are limited to planning, producing/updating plan documents (plans/YYYY-MM-DD-<slug>.md) using the provided template, asking clarifying questions, and enforcing approval/verification rules. This stays within the skill's purpose. One minor ambiguity: the phrase 'minimum files required to create or resume the plan' could be interpreted broadly (resuming a plan might involve reading project files for context); the SKILL.md otherwise discourages deep research and arbitrary file reads during the Plan phase.
Install Mechanism
There is no install spec embedded in the skill bundle (instruction-only). The README mentions optional git clone or ClawHub commands, but those are external installation suggestions, not part of the runtime skill content. No downloads or archive extraction are specified in the skill itself.
Credentials
The skill declares no required environment variables, credentials, or config paths. It does not ask for secrets. README instructions that reference git/ClawHub are normal for user-side installation and do not change the skill's runtime credential needs.
Persistence & Privilege
always is false and there is no code in the bundle that forces persistent autonomous behavior. The skill references background/sub-agent execution and progress pushing (platform features) — this is conceptually expected for a planning/execution skill, but users should be aware that actual background/sub-agent behavior depends on platform capabilities and permissions, not the skill bundle itself.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install pad-mode - 安装完成后，直接呼叫该 Skill 的名称或使用
/pad-mode触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.1.2
Add execution lease/watchdog and interruption recovery rules for PAD execution stability.
v1.2.0
pad-mode v1.2.0
- Clarifies that PAD Mode is a general execution framework, not just for coding; adds task-type awareness throughout.
- Expands and tightens enforcement guardrails (e.g., must verify tasks before marking done, avoid oversized tasks, require review for high-risk work).
- Adds “Definition of Done” guidance for multiple task types (coding, research, ops, content, general).
- Emphasizes use of small, trackable tasks (2–15 minutes) and explicit verification before marking any task complete.
- Enhances task breakdown and planning process to be sensitive to task type, risk, and verification needs.
- Minor editorial fixes and clarifications for phases and enforcement points.
v1.1.1
Improved README with badges, before/after comparison, enforcement table, real example, and trigger examples
v1.1.0
Added ⚠️ Enforcement rules (5 violation consequences), STOP gates in Phase 3/4/5 for approval enforcement
v1.0.8
Renamed display to PAD Mode
v1.0.7
Fix: bare /pad must ask user intent before creating plan, no context guessing
v1.0.6
Fixed: removed Chinese text from SKILL.md, English only
v1.0.5
Fixed: enforce Approve phase, button+text fallback for all channels, mandatory execution mode prompt
v1.0.4
Reverted display name
v1.0.3
Renamed to include 'Plan Mode' for search discoverability
v1.0.2
Added plan-mode keywords for better search discoverability
v1.0.1
Optimized description for better discoverability
v1.0.0
Initial release: Plan → Act → Deliver structured workflow for complex tasks
元数据
常见问题
Pad Mode 是什么？
Turn messy requests into structured plans. PAD Mode (Plan → Act → Deliver) gives your AI agent project management superpowers — automatic task breakdown, liv... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 286 次。
如何安装 Pad Mode？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install pad-mode」即可一键安装，无需额外配置。
Pad Mode 是免费的吗？
是的，Pad Mode 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Pad Mode 支持哪些平台？
Pad Mode 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 Pad Mode？
由 Yipxiyi（@yipxiyi）开发并维护，当前版本 v1.2.0。
推荐 Skills