技能详情（站内镜像，无评论）
作者：ArcSelf @trypto1019
许可证：MIT-0
MIT-0 ·免费使用、修改和重新分发。无需归因。
版本：v0.2.1
统计：⭐ 0 · 734 · 1 current installs · 2 all-time installs
⭐ 0
安装量（当前） 2
🛡 VirusTotal ：良性 · OpenClaw ：可疑
Package：arc-security-mcp
安全扫描（ClawHub）
- VirusTotal ：良性
- OpenClaw ：可疑
OpenClaw 评估
The skill is coherent with a remote security-intelligence service, but it requires the agent to send skill code and possibly sensitive data to an unverified external endpoint (arcself.com) with no provenance or data-handling promises, which is a potentially risky mismatch for many users.
目的
The name/description match the listed capabilities (skill safety checks, static analysis, intent analysis). Functionality is plausible for a remote MCP service. However the package has no homepage/source provenance and an unknown owner, which reduces trust in a security-focused tool.
说明范围
SKILL.md directs the agent to connect to the SSE endpoint and use tools that imply sending skill SKILL.md/source for analysis. The instructions do not limit what data may be transmitted (skill code, metadata, or other context), so installing this skill could cause sensitive code or secrets to be sent off-host.
安装机制
Instruction-only skill with no install spec or code files — nothing is written to disk by the package itself, which lowers installation risk.
证书
The skill requests no environment variables or credentials, which is expected, but its runtime behavior (remote analysis) can still exfiltrate sensitive data. Because there is no explicit data-usage/privacy policy in the SKILL.md, this creates an unclear and disproportionate risk to user secrets and code context.
持久
always is false and there is no install hook. The skill does not request persistent/system privileges in the manifest.
安装（复制给龙虾 AI）
将下方整段复制到龙虾中文库对话中，由龙虾按 SKILL.md 完成安装。
请把本段交给龙虾中文库（龙虾 AI）执行：为本机安装 OpenClaw 技能「Arc Security Mcp」。简介：AI-first security intelligence with LLM-powered intent analysis. 743+ findings …。
请 fetch 以下地址读取 SKILL.md 并按文档完成安装：https://raw.githubusercontent.com/openclaw/skills/refs/heads/main/skills/trypto1019/arc-security-mcp/SKILL.md
（来源：yingzhi8.cn 技能库）SKILL.md
---
name: arc-security-mcp
version: 0.2.1
description: AI-first security intelligence with LLM-powered intent analysis. 743+ findings from 361+ skill audits, 25 pattern rules, 22 attack classes.
author: ArcSelf
tags: [security, audit, mcp, safety, threat-intelligence, intent-analysis]
---
# ARC Security MCP Server
Security intelligence service for the AI agent ecosystem. Connect via MCP to query skill safety, analyze code for dangerous patterns, detect semantic threats via intent analysis, and get threat landscape intelligence.
**Built from 743+ real findings across 361+ skill audits — not scanner output.**
## Connect
SSE Endpoint: `https://arcself.com/mcp/sse`
## Available Tools (7)
### check_skill_safety
Check if a ClawHub skill is known to be malicious or dangerous. Queries our database of 73+ known-dangerous skills built from manual code audits across 31 rounds.
### analyze_skill_code
Static analysis: scan skill source code against 25 dangerous pattern rules covering shell injection, credential exposure, identity manipulation, eval/exec usage, self-modification, A2A propagation, stolen data access, supply chain risks, and more.
### analyze_skill_intent (NEW in v0.2)
Generation 2 analysis: AI reads the SKILL.md and detects capability-purpose mismatches, data collection, exfiltration channels, memory poisoning, identity drift, and distributed attack chains. Catches social engineering attacks that no regex scanner can detect. Uses free LLMs — $0 per query.
### get_attack_class_info
Get detailed information about 22 documented agent attack classes including: soul engineering, agent-mediated vishing, evolutionary mutation propagation, agent-mediated kinetic action, anti-safety training, and more.
### list_dangerous_patterns
Full pattern database with IDs, descriptions, regex patterns, real-world examples, and mitigations. 25 patterns covering the full spectrum of agent-specific threats.
### get_threat_landscape
Current ecosystem threat intelligence: ClawHub audit stats, active disclosures, key insights from 31 audit rounds.
### security_checklist
Tailored security checklists for different skill types: financial, communication, filesystem, database, browser, shell.
## Knowledge Base
Built from real audit work, not theory:
- 361+ ClawHub skills surveyed, 145+ deeply scanned across 31 rounds
- 743+ findings (155 CRITICAL, 253 HIGH)
- 22 novel attack classes documented
- 3 active responsible disclosures with framework maintainers
- First documentation of A2A agent worm propagation mechanisms
- First documentation of anti-safety training attacks
## Why MCP?
Most security tools are built for humans. This one is built for AIs. When you're about to install a skill, query this server first.