← 返回 Skills 市场
Sentinel Vanguard AI Skill Security Auditor
作者
DTTNpole-commits
· GitHub ↗
· v2.0.1
· MIT-0
142
总下载
0
收藏
0
当前安装
3
版本数
在 OpenClaw 中安装
/install sentinel-vanguard
功能描述
AI Agent skill security auditor. Use this skill whenever the user wants to audit, review, vet, or assess the safety and security of any AI skill, Claude skil...
安全使用建议
This skill is low-risk because it is instruction-only and explicitly forbids network access, code execution, and credential access. Before using it: (1) avoid pasting any sensitive secrets or private keys into the audit input (the skill analyzes whatever you paste); (2) remember the auditor relies on pattern matching and heuristics — its blocklist and rules may be out of date or produce false negatives/positives; and (3) because it runs as an instruction set inside your agent, ensure you trust the agent runtime (the skill itself will not fetch remote content or install software).
功能分析
Type: OpenClaw Skill
Name: sentinel-vanguard
Version: 2.0.1
The 'sentinel-vanguard' skill is a security auditing tool designed to perform static, logic, and supply-chain analysis on other AI agent skills. The instructions in SKILL.md explicitly enforce a read-only, no-network, and no-code-execution policy, while the reference files (l1-rules.md and l3-blocklist.md) provide legitimate security signatures and known-malicious package lists to assist the agent in identifying risks. There is no evidence of malicious intent, data exfiltration, or adversarial instructions.
能力评估
Purpose & Capability
The skill is an auditor and is instruction-only; it requests no binaries, env vars, or installs and its internal rule lists and scoring align with a static/text-based audit use case.
Instruction Scope
SKILL.md explicitly constrains the agent to read-only text analysis of user-pasted content, forbids network calls, code execution, and secret access; the three-layer scans operate only on provided text and reference only pattern/heuristic checks.
Install Mechanism
There is no install spec and no code files; the skill is purely instruction-driven so there is no code to be written to disk or downloaded at install-time.
Credentials
The skill declares no required environment variables, no primary credential, and the instructions explicitly forbid reading secrets or environment state; requested privileges are minimal and appropriate.
Persistence & Privilege
always is false, it does not request permanent presence or modification of other skills, and there are no instructions to change agent/system configuration.
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 sentinel-vanguard - 安装完成后，直接呼叫该 Skill 的名称或使用
/sentinel-vanguard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
v2.0.1
**Summary:** v2.0.1 introduces strict read-only, text-only auditing; bans network/code execution; and requires users to paste skill content directly.
- Added hard constraints: No network requests, no code execution, no credential access—analysis is now strictly read-only and text-based.
- Changed workflow: Skill only accepts pasted content; instructs users to paste skill/code instead of submitting URLs.
- Updated report and protocol descriptions to emphasize that only direct text input is analyzed.
- Clarified and simplified accepted input types and reporting format.
- Tightened behavioral rules for improved security and clarity.
v2.0.0
- Clarified and expanded audit protocol descriptions for all three layers, improving guidance and risk definitions.
- Refined L1 static scan and L2 logic analysis categories with more detailed rule language and explicit examples.
- Updated report output format to be more concise, removing Markdown header formatting requirements and simplifying table structures.
- Standardized the terminology for risk bands, permission matrix entries, and dependency audit criteria.
- Improved behavioural rules and explanations, ensuring consistent and actionable audit recommendations for all users.
v1.0.0
Sentinel Vanguard 1.0.0 — Initial release
- Introduces a three-layer security audit protocol for AI agent skills: static scan, logic (prompt injection) analysis, and supply chain (dependency) audit.
- Detects destructive operations, exfiltration behavior, dangerous code patterns, and permission anomalies.
- Automatically analyzes prompts and instructions for direct, indirect, and jailbreak-oriented prompt injection vectors.
- Audits skill dependencies for known malicious packages, typosquatting, and unsafe versioning.
- Generates a detailed, standardized Markdown security report with clear verdicts, permission matrix, findings, and remediation checklist.
- Designed for integration with ClawHub and SkillPay platforms and accepts multiple input formats (code, URLs, metadata).
元数据
常见问题
Sentinel Vanguard AI Skill Security Auditor 是什么？
AI Agent skill security auditor. Use this skill whenever the user wants to audit, review, vet, or assess the safety and security of any AI skill, Claude skil... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 142 次。
如何安装 Sentinel Vanguard AI Skill Security Auditor？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install sentinel-vanguard」即可一键安装，无需额外配置。
Sentinel Vanguard AI Skill Security Auditor 是免费的吗？
是的，Sentinel Vanguard AI Skill Security Auditor 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Sentinel Vanguard AI Skill Security Auditor 支持哪些平台？
Sentinel Vanguard AI Skill Security Auditor 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 Sentinel Vanguard AI Skill Security Auditor？
由 DTTNpole-commits（@dttnpole-commits）开发并维护，当前版本 v2.0.1。
推荐 Skills