← Back to Skills Marketplace
20 Agent Security Questions
by
caidongyun
· GitHub ↗
· v1.0.0
· MIT-0
262
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install 20-agent-security-questions
Description
Automation skill for 20 Agent Security Questions.
README (SKILL.md)
name: agent-security-questions description: 工作问题收集与分析Skill - 收集AI Agent工作过程中的问题，分析风险，转化为研究课题 metadata: openclaw: emoji: ❓ version: 1.0.0
❓ agent-security-questions
工作问题收集与分析Skill
功能
- 问题收集 - 收集工作过程中的问题
- 分类分析 - 分析问题类型和风险
- 研究转化 - 将问题转化为研究课题
- 知识沉淀 - 沉淀为知识库
问题类型
| 类型 | 说明 |
|---|---|
| 工作问题 | Agent工作过程中的问题 |
| 安全风险 | 安全相关风险 |
| 机制问题 | 智能体机制问题 |
| 产出问题 | 产出质量问题 |
| 沟通问题 | 人机协作问题 |
使用
./src/questions.sh add "问题描述"
./src/questions.sh list
./src/questions.sh analyze
Usage Guidance
This package appears to be a stub: the README promises collection and analysis, but the only code is a placeholder script that prints usage. It's low-risk now (no network calls or credential requests), but it's functionally incomplete. Before installing or enabling it for autonomous use, ask the author for the real implementation or inspect any future updates closely. Run the skill in a sandbox to verify behavior, and avoid granting any credentials or elevated permissions unless the skill's full, concrete behavior is available and matches its description.
Capability Analysis
Type: OpenClaw Skill
Name: 20-agent-security-questions
Version: 1.0.0
The skill bundle is a non-functional template or placeholder for tracking AI agent security questions. The shell script src/questions.sh is a simple stub that only prints usage instructions, and the SKILL.md documentation describes standard problem-tracking tasks without any malicious instructions, data exfiltration, or risky execution patterns.
Capability Assessment
Purpose & Capability
The SKILL.md describes collecting, classifying, analyzing, and converting problems to research topics, but the repository contains only a tiny src/questions.sh that prints a usage message; there is no implementation of the stated capabilities or the advertised '20 questions'. This is a functional mismatch (likely unfinished or a stub), not an explained feature.
Instruction Scope
Runtime instructions tell the agent to run ./src/questions.sh add|list|analyze, but the script doesn't implement those actions and does not read files, network, or credentials. The instructions themselves don't instruct any data exfiltration, but they promise behaviors that are not present in the code.
Install Mechanism
No install spec and only a small local script are included. There is no downloading, extracting, or third‑party package installation; risk from installation mechanism is minimal.
Credentials
The skill declares no required environment variables, no credentials, and no config paths. There is nothing requesting unrelated secrets or access.
Persistence & Privilege
The skill does not request always: true, does not modify other skills, and has default autonomous invocation settings. It has no apparent persistence or elevated privileges.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install 20-agent-security-questions - After installation, invoke the skill by name or use
/20-agent-security-questions - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
Initial release of agent-security-questions skill.
- Collects issues encountered during AI agent operations.
- Analyzes and categorizes problems by risk type.
- Converts identified issues into research topics.
- Builds a knowledge base for ongoing insights.
- Provides command-line scripts for adding, listing, and analyzing questions.
Metadata
Frequently Asked Questions
What is 20 Agent Security Questions?
Automation skill for 20 Agent Security Questions. It is an AI Agent Skill for Claude Code / OpenClaw, with 262 downloads so far.
How do I install 20 Agent Security Questions?
Run "/install 20-agent-security-questions" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is 20 Agent Security Questions free?
Yes, 20 Agent Security Questions is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does 20 Agent Security Questions support?
20 Agent Security Questions is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created 20 Agent Security Questions?
It is built and maintained by caidongyun (@caidongyun); the current version is v1.0.0.
More Skills