360 Secure SkillHub
360 安全技能中心

Kefal Guard

面向 OpenClaw 网关的只读主机安全监控 Skill:需手动安装 kefal-agent 二进制,定时扫描进程、监听端口、账户与 SSH 公钥指纹,检测暴露服务、提权路径和未知外连,提供事件列表、严重级别、修复命令与基础设施图谱;遥测经 TLS 上传至 Kefal 云端仪表盘。

davidangularme
davidangularme
热度 33.84
最近更新 2026-05-11 17:45:50
分类:安全依赖:代码执行 / 联网浏览搜索 / 第三方 API应用场景:DevOps / 报告撰写 / 自动化脚本

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add kefal-guard
Metadata
name
kefal_guard
description
Infrastructure security monitor — detects exposed services, privilege escalation paths, and novel threats using compositional reasoning. Read-only host telemetry agent. Manual install, no auto-execution.
metadata
{"openclaw":{"os":["darwin","linux"],"requires":{"bins":["kefal-agent"]},"homepage":"https://kefal.dev","docs":"https://kefal.dev/docs/installation.html"}}

Kefal Guard — Infrastructure Security for OpenClaw

Kefal monitors the machine running your OpenClaw gateway. It detects exposed databases, privilege escalation paths, unauthorized SSH keys, novel outbound connections, and services listening where they shouldn't — before an attacker does.

What it does

Kefal installs a lightweight agent (single static binary, ~6 MB) that scans your system every 60 seconds. It reports to your Kefal dashboard at kefal.dev/app, where you see:

  • A live graph of every host, service, port, and identity on your infrastructure
  • Incidents with severity levels (critical, high, medium) and plain-English explanations
  • Remediation plans with the exact shell commands to fix each issue

32 security invariants run on every scan. When something breaks — a new admin account, a service exposed to the internet, an outbound connection to an unknown IP — Kefal tells you what happened, why it matters, and how to fix it.

When to use this skill

  • After installing OpenClaw on a new server or VPS
  • Before exposing your gateway to the internet
  • When installing untrusted third-party skills from ClawHub
  • As continuous monitoring alongside your OpenClaw agent

Install

The agent is a separate binary that the operator installs and reviews manually — this skill does not download or execute remote code. Platform-specific install instructions, binary checksums, and verification steps are documented at:

https://kefal.dev/docs/installation.html

After install, verify with kefal-agent --version and kefal-agent --status. The skill below assumes the binary is on $PATH.

Commands

  • /kefal status — Show agent status, last scan time, and open incidents
  • /kefal scan — Trigger an immediate scan (normally runs every 60s)
  • /kefal incidents — List all open incidents with severity and details
  • /kefal remediate <incident-id> — Show remediation steps for a specific incident
  • /kefal graph — Open the live infrastructure graph in your browser

How the agent uses this skill

When the user asks about security, infrastructure health, or mentions concerns about exposed services or vulnerabilities, use the exec tool to run kefal-agent --status or kefal-agent --scan and report the results. If incidents are found, present them with their remediation steps.

When the user installs a new third-party skill, suggest running /kefal scan to verify no new attack surface was introduced.

When the user asks "is my server secure?" or "audit my setup", run a full scan and present the results as a security report.

What data does the agent collect

The agent reads host system metadata only:

  • Running processes — name, PID, owning user, CPU usage. (Source: /proc/<pid>/stat on Linux, equivalent APIs on macOS/Windows. The agent does not read process memory or open file descriptors.)
  • Listening ports — port number, protocol, owning process name. (Source: /proc/net/tcp and equivalents. The agent does not capture packets or read network traffic.)
  • User accounts with shell access — username and login state. (Source: /etc/passwd for the username list, utmp for active sessions. The agent reads these as a non-root user when possible; sudo is only needed for full process visibility.)
  • SSH authorized keys — public key fingerprints in ~/.ssh/authorized_keys. (Source: the file is read line-by-line as text. The agent never touches private keys; private keys live in different files and are never read.)

The agent does NOT read application data (databases, application config, business documents), does NOT capture network traffic, and does NOT modify any system file. All telemetry is transmitted over TLS 1.3 to kefal.dev. Each tenant's data is isolated; no data is shared with other customers.

The agent source structure is documented in the installation guide and the binaries are published with reproducible-build flags (-trimpath -ldflags="-s -w"), so the SHA-256 you download can be matched against a future open-source release.

Pricing

The skill includes a 7-day free trial. Plans start at $49/month for up to 3 agents. No credit card required to start. See https://kefal.dev/#pricing for details.

Built by

Catalyst AI Research — Haifa, Israel. https://catalystais.com