← 返回 Skills 市场
CRUD
作者
Slava Chan
· GitHub ↗
· v1.0.3
· MIT-0
157
总下载
0
收藏
0
当前安装
4
版本数
在 OpenClaw 中安装
/install crud
功能描述
管理OpenClaw操作按CRUD分类，创建查询无审批，更新编辑删除需返回操作清单并二次确认。
使用说明 (SKILL.md)
CRUD Approval Mode Skill
Categorize and manage OpenClaw operations by CRUD (Create/Read/Update/Delete) for fine-grained approval control.
Core Design Philosophy
- Read: Fully unrestricted for maximum efficiency
- Create + Update + Edit + Delete: Add an isolation layer, return action list for secondary confirmation
Category Definitions
| Operation Type | Approval Required | Description |
|---|---|---|
| Read | ❌ No approval | File reading, search queries, directory browsing, etc. |
| Create | ✅ Requires confirmation | File creation, new skill creation, directory creation, etc. |
| Update/Edit | ✅ Requires confirmation | File modification, code editing, config updates, etc. |
| Delete | ✅ Requires confirmation | File deletion, directory deletion, skill deletion, etc. |
Workflow
1. No Approval Flow (Read)
User Request → Execute Directly → Return Result
2. Approval Flow (Create/Update/Edit/Delete)
User Request → Analyze Operation → Generate Action List → User Confirmation → Execute Operation → Return Result
↓
Display: operation type, affected files, change preview
Action List Format
When confirmation is needed, return the list in the following format:
⚠️ **Operation Confirmation** - [Operation Type]
**Scope:**
- File 1: path/to/file1
- File 2: path/to/file2
**Description:**
[Detailed description of the operation to be performed]
Please reply with one of the following to continue:
- ✅ **Confirm** - Proceed with this operation
- ❌ **Cancel** - Cancel this operation
- 🔄 **Modify** - Modify operation and re-confirm
Use Cases
- Daily development: Quick read operations without approval restrictions
- Critical operations: Confirmation mechanism before all write operations
- Skill development: New skill creation requires confirmation for safety
Configuration
This skill requires no additional configuration files and is automatically injected into sessions.
References
- OpenClaw Documentation: https://docs.openclaw.ai/
安全使用建议
This skill appears to do what it claims: classify operations and require confirmation for writes. Before installing: 1) Be aware that 'Read' operations are allowed without confirmation — in sensitive environments you may prefer a stricter policy or to limit the skill's ability to read certain paths. 2) Verify how your OpenClaw instance loads/injects skills (the docs in the package say it's 'automatically injected' which conflicts with the manifest flags); ensure it won't be auto-enabled in contexts you don't expect. 3) Test in a non-production environment to confirm the confirmation flow works and that action lists accurately list affected files. 4) Use the provided uninstall steps if you need to remove it. 5) Only install from a trusted source; this package has no homepage or verified publisher metadata, so validate the origin if you rely on it in production.
功能分析
Type: OpenClaw Skill
Name: crud
Version: 1.0.3
The 'crud' skill is a governance tool designed to enforce a human-in-the-loop approval process for state-changing operations (Create, Update, Delete) while allowing Read operations freely. The bundle consists entirely of Markdown instructions (SKILL.md, CLAUDE.md) and documentation (README.md, EXAMPLES.md) that guide the AI agent's behavior to improve safety; it contains no executable code, suspicious network calls, or evidence of malicious intent.
能力评估
Purpose & Capability
The name/description (CRUD approval gating) match the SKILL.md, CLAUDE.md, README and examples. The skill declares no env vars, binaries, or installs, which is reasonable for an instruction-only governance/workflow skill.
Instruction Scope
Instructions are narrowly focused on classifying operations and requiring user confirmation for Create/Update/Edit/Delete. However, the policy explicitly allows Read operations to execute immediately (file reads, directory listing, search) with no approval. That is consistent with the stated design but grants broad read access which could surface sensitive data; the skill gives no guidance on limiting read scope or filtering sensitive files.
Install Mechanism
This is an instruction-only skill with no install spec and no code to fetch or execute. That minimizes installation risk.
Credentials
The skill requests no environment variables, credentials, or special config paths — consistent and proportionate for a policy/approval skill.
Persistence & Privilege
Metadata flags show no 'always' privilege (always: false) and normal autonomous invocation is allowed. The documentation text says the skill is 'automatically injected into sessions,' which is misleading relative to the flags; confirm how your OpenClaw deployment will load/inject the skill before trusting automatic behavior. The skill does not request system-wide privileges or modify other skills.
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 crud - 安装完成后，直接呼叫该 Skill 的名称或使用
/crud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
v1.0.3
- Updated author information in metadata (now lists "Slava Chan @UyNewNas").
- No other content or functional changes.
v1.0.2
Version 2.0.0
- Major change: Approvals are now required for Create operations in addition to Update/Edit/Delete.
- Only Read operations remain fully unrestricted; all write-type actions (Create/Update/Edit/Delete) now require user confirmation.
- SKILL.md updated to reflect new approval flows, category definitions, and use cases.
- Removed prior exception for unrestricted Create; new skills and file creations are now subject to confirmation.
v1.0.1
- English translation of documentation; all descriptions and guides are now in English.
- Added link to an external reference project in the References section.
- No code or configurable behavior has changed (documentation update only).
v1.0.0
Further subdivide OC(OpenClaw) behaviors into CRUD-based classifications:
• For Create and Read operations, unrestricted access is allowed, though newly created skills may require manual review after creation.
• For Update, Edit, and Delete operations, an additional isolation layer is added. During the conversation, a behavior checklist is returned, and a unified secondary confirmation is performed within the dialogue. This aligns with typical SaaS software behavior, emphasizing focused confirmation.
元数据
常见问题
CRUD 是什么？
管理OpenClaw操作按CRUD分类，创建查询无审批，更新编辑删除需返回操作清单并二次确认。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 157 次。
如何安装 CRUD？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install crud」即可一键安装，无需额外配置。
CRUD 是免费的吗？
是的，CRUD 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
CRUD 支持哪些平台？
CRUD 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 CRUD？
由 Slava Chan（@uynewnas）开发并维护，当前版本 v1.0.3。
推荐 Skills