中文 EN
开始可信审查
安全决策报告

memphis-cognitive

Documentation-only skill that prominently instructs users to execute remote curl|bash installers without adequate warnings, enabling supply chain attack vectors through untrusted external scripts.

安装决策优先 来源: ClawHub 扫描时间: 2026/6/24
文件 6
IOC 8
越权项 0
发现 4
最直接的威胁证据
严重 供应链
Remote script execution via curl|bash

SKILL.md, QUICKSTART.md, and README.md instruct users to execute remote shell scripts directly from GitHub raw URLs without verification. This is a primary supply chain attack vector.

QUICKSTART.md:13

为什么得出这个结论

2/4 个维度触发
通过
声明与实际能力

声明资源与推断能力基本一致。

阻止
隐藏执行与外联

提取到 2 个高危 IOC 或外联信号。

阻止
攻击链与高危发现

报告包含 4 步攻击链，另有 2 项高危或严重发现。

复核
依赖与供应链卫生

没有完整依赖信息，供应链判断需要保留弹性。

攻击链

01
User follows skill documentation

initial_access · QUICKSTART.md:13

02
Executes curl|bash from unverified GitHub raw URL

delivery · QUICKSTART.md:13

03
install.sh runs arbitrary code with user/root privileges

代码执行 · https://raw.githubusercontent.com/elathoxu-crypto/memphis/main/install.sh:1

04
Malicious code in install.sh could: install backdoor, exfiltrate credentials, establish persistence

最终危害 · external

风险分是怎么被拉高的

Remote curl|bash execution in documentation +35

QUICKSTART.md:13 and README.md:128 instruct users to pipe remote scripts directly to bash/shell

No declared permissions vs undeclared shell execution vectors +15

Skill declares NONE permissions but guides users to execute arbitrary remote code

Opaque third-party dependency +10

Actual functionality lives in externally-installed Memphis CLI from unverified GitHub repo elathoxu-crypto/memphis

Known risk disclosure in skill-card.md +-10

Risk acknowledged in skill-card.md but not prominently surfaced in SKILL.md itself

最关键的证据

严重 供应链

Remote script execution via curl|bash

SKILL.md, QUICKSTART.md, and README.md instruct users to execute remote shell scripts directly from GitHub raw URLs without verification. This is a primary supply chain attack vector.

QUICKSTART.md:13
Remove curl|bash one-liner. Use manual installation steps with version-pinned releases or require users to inspect scripts before execution.
严重 供应链

Second remote script execution via curl|sh

README.md instructs users to install Ollama via curl|sh remote script, compounding supply chain risk.

README.md:128
Remove curl|sh for Ollama. Point to official package manager instructions or require manual inspection.
中危 文档欺骗

Security risks not prominently surfaced in SKILL.md

skill-card.md contains explicit risk disclosures about privileged shell installers and sensitive data handling, but SKILL.md—the primary documentation users interact with—does not prominently warn about these dangers.

SKILL.md:1
Add prominent security section at top of SKILL.md warning about remote script risks and data sensitivity.
低危 权限提升

Potential privilege escalation through npm install -g

Installation instructions suggest npm install -g which may require sudo privileges, executing with elevated permissions.

README.md:62
Document that global npm installs may require privilege escalation and advise caution.

声明能力 vs 实际能力

文件系统 通过
声明 NONE
→
推断 NONE
No file writes in skill files; wrapper only checks CLI presence
命令执行 通过
声明 NONE
→
推断 NONE
Wrapper uses exec memphis $@ only; no direct shell execution in skill
网络访问 通过
声明 NONE
→
推断 NONE
Skill files do not make network requests
credential 通过
声明 NONE
→
推断 NONE
No credential access in skill code
环境变量 通过
声明 NONE
→
推断 NONE
No env access in skill files

可疑产物与外联

严重 危险命令
curl -fsSL https://raw.githubusercontent.com/elathoxu-crypto/memphis/main/install.sh | bash

QUICKSTART.md:13

严重 危险命令
curl -fsSL https://ollama.com/install.sh | sh

README.md:128

中危 外部 URL
https://deb.nodesource.com/setup_20.x

README.md:90

中危 外部 URL
https://ollama.com/install.sh

README.md:128

中危 外部 URL
https://clawhub.com/skill/memphis-cognitive

README.md:268

中危 外部 URL
https://discord.gg/clawd

README.md:269

中危 外部 URL
https://clawhub.ai/user/elathoxu-crypto

skill-card.md:7

中危 外部 URL
https://clawhub.ai/elathoxu-crypto/memphis-cognitive

skill-card.md:27

依赖与供应链

没有结构化依赖告警。

文件构成

6 个文件 · 983 行
Markdown 4 个文件 · 961 行Shell 1 个文件 · 17 行JSON 1 个文件 · 5 行
需关注文件 · 4
SKILL.md Markdown · 536 行
Security risks not prominently surfaced in SKILL.md
README.md Markdown · 309 行
Second remote script execution via curl|sh · Potential privilege escalation through npm install -g · curl -fsSL https://ollama.com/install.sh | sh · https://deb.nodesource.com/setup_20.x · https://ollama.com/install.sh · https://clawhub.com/skill/memphis-cognitive · https://discord.gg/clawd
skill-card.md Markdown · 43 行
https://clawhub.ai/user/elathoxu-crypto · https://clawhub.ai/elathoxu-crypto/memphis-cognitive
QUICKSTART.md Markdown · 73 行
Remote script execution via curl|bash · curl -fsSL https://raw.githubusercontent.com/elathoxu-crypto/memphis/main/install.sh | bash
其他文件 · memphis-wrapper.sh · _meta.json

安全亮点

Skill is explicitly documented as documentation-only (META-PACKAGE), no hidden compiled code
Wrapper script (memphis-wrapper.sh) is simple and benign - only checks CLI presence and forwards args
skill-card.md acknowledges supply chain risks and advises users to review scripts before execution
No credential theft, data exfiltration, or obfuscation observed in skill files themselves
No base64-encoded execution, reverse shells, or C2 communications in skill code
MIT license, open source repository declared
Local-first architecture (no mandatory cloud dependency for core functions)