中文 EN

Scan Report

Scan New Files

✓ Low Risk — Risk Score 8/100
Last scan:1 day ago Rescan
8 /100
gpt-chat
GPT聊天 - 使用GPT模型进行对话和内容生成
Legitimate GPT chat skill with declared network/environment access and no malicious behavior; minor doc-to-config mismatch on the default API proxy.
Skill Namegpt-chat
Duration35.7s
Enginepi
✓
Safe to install
Consider documenting the default api.openai-proxy.org proxy and pinning node version in package.json for reproducibility.

Findings 2 items

Severity Finding Location
Low
Undeclared default API proxy endpoint Doc Mismatch
The skill's main script gpt.js defaults to api.openai-proxy.org as the API_BASE instead of the official api.openai.com, but SKILL.md only mentions OPENAI_API_KEY with no mention of the proxy or proxy configuration. The proxy URL can be overridden via OPENAI_API_BASE env var, but this is not documented.
const API_BASE = process.env.OPENAI_API_BASE || 'https://api.openai-proxy.org';
→ Document the OPENAI_API_BASE environment variable and the default proxy in SKILL.md, or default to api.openai.com for transparency.
 scripts/gpt.js:11
Low
Non-existent GPT-5 model references Doc Mismatch
The skill advertises GPT-5.2, GPT-5.1, and GPT-5 models with pricing tables in SKILL.md and code constants, but these models do not exist on the public OpenAI API. This is misleading documentation for a model-selector skill.
| GPT-5.2 | $1.75/1M tokens | $14/1M tokens | 最新旗舰模型 |
→ Use actual OpenAI model names (gpt-4o, gpt-4o-mini, gpt-4-turbo, etc.) or clearly note that these are fictional/demonstration models.
 SKILL.md:16
ResourceDeclaredInferredStatusEvidence
Filesystem NONE READ+WRITE ✓ Aligned SKILL.md:1 — script execution implies node access; gpt.js:state.json write; gene…
Network READ READ ✓ Aligned gpt.js:36-62 — HTTPS POST to api.openai-proxy.org (or env-configured endpoint)
Environment READ READ ✓ Aligned gpt.js:11 reads OPENAI_API_KEY and OPENAI_API_BASE from process.env
Shell NONE NONE — No shell/shutdown/exec calls; server.js uses Node.js process signals which are s…
Skill Invoke NONE NONE — No cross-skill invocation found
Clipboard NONE NONE — Not accessed
Browser NONE NONE — Not accessed
Database NONE NONE — Not accessed
1 findings
🔗
Medium External URL 外部 URL
https://api.openai-proxy.org
scripts/gpt.js:11

File Tree

9 files · 14.1 KB · 530 lines
JavaScript 6f · 455L Markdown 1f · 60L JSON 2f · 15L
├─ ▾ 📁 scripts
│ ├─ 📜 generate-article.js JavaScript 73L · 1.9 KB
│ ├─ 📜 gpt.js JavaScript 136L · 3.9 KB
│ ├─ 📜 server.js JavaScript 130L · 3.6 KB
│ ├─ 📋 state.json JSON 2L · 31 B
│ ├─ 📜 test-chat.js JavaScript 49L · 1.2 KB
│ ├─ 📜 test-key.js JavaScript 31L · 817 B
│ └─ 📜 test.js JavaScript 36L · 754 B
├─ 📋 package.json JSON 13L · 338 B
└─ 📝 SKILL.md Markdown 60L · 1.5 KB

Security Positives

✓ No shell execution, reverse shells, or command injection patterns detected
✓ No credential harvesting beyond the declared OPENAI_API_KEY
✓ No base64-encoded payloads, eval(), or obfuscation techniques
✓ No sensitive file path access (~/.ssh, ~/.aws, .env) beyond state.json
✓ No data exfiltration or C2 communication patterns
✓ No supply chain risks — package.json has no external dependencies
✓ HTTPS used for all API calls with Bearer token authentication
✓ State persistence is local-only (state.json in scripts directory)