← 返回 Skills 市场
107
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install plan-first
功能描述
Solve complex multi-step tasks by generating a detailed plan before execution. Based on the Plan-and-Solve prompting research, this skill breaks tasks into c...
安全使用建议
This skill is coherent and low-risk as-distributed: it's a prompt/template for planning complex tasks and does not by itself access your system or secrets. Before using it in an agent that will execute steps autonomously, review any generated plan carefully and do not supply credentials unless you intend the agent to perform actions that require them. If you want to limit risk, keep this skill user-invocable only (avoid allowing fully autonomous runs), and require human approval before executing any plan steps that perform network calls, modify databases, or run shell commands.
功能分析
Type: OpenClaw Skill
Name: plan-first
Version: 1.0.0
The 'plan-first' skill bundle is an instructional framework designed to improve AI agent reasoning through structured planning and step-by-step execution. It consists entirely of Markdown documentation (SKILL.md and references/examples.md) providing templates and examples for task management; it contains no executable code, no network activity, and no malicious instructions or prompt-injection attacks.
能力评估
Purpose & Capability
The name and description match the SKILL.md: the skill teaches a plan-then-execute workflow and provides templates and examples. It declares no binaries, env vars, or installs, which is appropriate for a purely prompting/instructional skill.
Instruction Scope
The runtime instructions are planning-focused and do not tell the agent to read system files, exfiltrate data, or call external endpoints. Examples reference real-world actions (DB migrations, sending email, running CLI tools) — which is expected — but are inherently open-ended: if the agent later executes a plan it generates, that execution could require credentials or run commands. The skill itself does not include execution commands or hidden I/O instructions.
Install Mechanism
No install spec and no code files. Lowest-risk class: nothing is written to disk or downloaded by the skill.
Credentials
No environment variables, credentials, or config paths are requested. Example plans mention services (SMTP, databases, cloud tools) but the skill does not request unrelated secrets or persistent credentials.
Persistence & Privilege
always is false and the skill does not request elevated or persistent privileges. disable-model-invocation is false (the platform default) but this is not combined with other risky indicators here.
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 plan-first - 安装完成后，直接呼叫该 Skill 的名称或使用
/plan-first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
v1.0.0
- Initial release of the plan-first skill, providing a structured, step-by-step approach for solving complex multi-step tasks.
- Implements Plan-and-Solve research methods: always generate and validate a detailed plan before taking any action.
- Includes a markdown plan template with dependencies, verification, files/tools, risk analysis, and rollback sections.
- Contains clear guidance on when (and when not) to apply plan-first methodology.
- Offers practical examples for coding, writing, and analysis tasks to illustrate effective usage.
元数据
常见问题
Plan First 是什么？
Solve complex multi-step tasks by generating a detailed plan before execution. Based on the Plan-and-Solve prompting research, this skill breaks tasks into c... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 107 次。
如何安装 Plan First？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install plan-first」即可一键安装，无需额外配置。
Plan First 是免费的吗？
是的，Plan First 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Plan First 支持哪些平台？
Plan First 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 Plan First？
由 Simon（@simoncatbot）开发并维护，当前版本 v1.0.0。
推荐 Skills