安全决策报告

gpt-chat

Multiple undocumented capabilities found: HTTP server with undeclared endpoints, undeclared filesystem writes for state persistence and file generation, and undeclared environment variable usage.

安装决策优先 来源: ClawHub 扫描时间: 17 天前
文件 11
IOC 3
越权项 1
发现 5
最直接的威胁证据

为什么得出这个结论

1/4 个维度触发
阻止
声明与实际能力

发现 1 项声明之外的能力或越权行为。

复核
隐藏执行与外联

提取到 3 个一般风险产物,需要结合上下文判断。

通过
攻击链与高危发现

没有形成明确的恶意路径。

通过
依赖与供应链卫生

依赖结构存在,但暂未看到明显高危告警。

风险分是怎么被拉高的

Undeclared HTTP server +15

server.js creates HTTP listener on port 3456 with /chat, /set-model endpoints - not mentioned in SKILL.md

Undeclared filesystem writes +10

state.json persistence and article file generation not declared

Undeclared environment variable +5

OPENAI_API_BASE used but not documented in requires section

Test scripts included +5

test.js, test-key.js, test-chat.js present but not referenced in documentation

External proxy usage +5

api.openai-proxy.org used as default instead of official OpenAI API

No malicious behavior confirmed +0

All code appears to be legitimate GPT API wrapper functionality

最关键的证据

中危 文档欺骗

Undeclared HTTP server with API endpoints

server.js creates an HTTP server on port 3456 that exposes /health, /set-model, /chat, and /models endpoints. This network listener functionality is completely absent from SKILL.md.

scripts/server.js:55
Document the HTTP server capability if intentional, or remove server.js from the skill package.
中危 文档欺骗

Undeclared filesystem write operations

gpt.js reads and writes state.json for model persistence, and generate-article.js writes output files. Neither is declared in SKILL.md.

scripts/gpt.js:32
Add filesystem:WRITE to declared capabilities if state persistence is intentional.
低危 文档欺骗

Undeclared environment variable usage

OPENAI_API_BASE environment variable is used as the API endpoint but is not listed in the requires.env section of SKILL.md.

scripts/gpt.js:10
Document all required and optional environment variables in SKILL.md.
低危 文档欺骗

Undocumented test scripts

test.js, test-key.js, and test-chat.js exist but are not referenced anywhere in SKILL.md documentation.

scripts/test.js:1
Either document these helper scripts or remove them from the skill package.
低危 供应链

External proxy dependency

Skill defaults to api.openai-proxy.org instead of official OpenAI API endpoints, introducing an untrusted intermediary.

scripts/gpt.js:11
Consider documenting the proxy requirement or providing official OpenAI endpoints as default.

声明能力 vs 实际能力

网络访问 通过
声明 READ
推断 READ+WRITE
server.js:55 creates HTTP server that accepts POST requests
文件系统 阻止
声明 NONE
推断 WRITE
gpt.js:32 writes state.json; generate-article.js:68 writes ai-future-article.md
环境变量 通过
声明 READ
推断 READ
OPENAI_API_KEY read in all scripts
命令执行 通过
声明 NONE
推断 NONE
No subprocess or shell execution found

可疑产物与外联

中危 外部 URL
https://api.openai-proxy.org

scripts/gpt.js:11

中危 外部 URL
https://clawhub.ai/user/lilisidu1210-ui

skill-card.md:7

中危 外部 URL
https://clawhub.ai/lilisidu1210-ui/gpt-chat

skill-card.md:28

依赖与供应链

包名版本来源漏洞备注
node system system No npm dependencies - uses only built-in Node.js modules (fs, https, http, url)
api.openai-proxy.org N/A external External proxy service used as default API endpoint

文件构成

11 个文件 · 576 行
JavaScript 6 个文件 · 455 行Markdown 2 个文件 · 101 行JSON 3 个文件 · 20 行
需关注文件 · 4
scripts/gpt.js JavaScript · 136 行
Undeclared filesystem write operations · Undeclared environment variable usage · External proxy dependency · https://api.openai-proxy.org
scripts/server.js JavaScript · 130 行
Undeclared HTTP server with API endpoints
skill-card.md Markdown · 41 行
https://clawhub.ai/user/lilisidu1210-ui · https://clawhub.ai/lilisidu1210-ui/gpt-chat
scripts/test.js JavaScript · 36 行
Undocumented test scripts
其他文件 · generate-article.js · SKILL.md · test-chat.js · test-key.js · package.json · _meta.json +1

安全亮点

No credential harvesting or exfiltration detected
No reverse shell or remote code execution patterns found
No base64 encoding or obfuscation detected
No access to sensitive paths like ~/.ssh or .env files
API key is only used for legitimate OpenAI API calls
No cron jobs, backdoors, or persistence mechanisms found
No suspicious network connections to unknown IPs