中文 EN

扫描报告

扫描新文件

✓ 可信 — 风险评分 5/100
上次扫描：2 天前 重新扫描
5 /100
neural-memory
Persistent memory for AI agents — stores experiences as interconnected neurons, recalls through spreading activation
Pure documentation skill package containing only SKILL.md files for NeuralMemory operations with no executable code, no sensitive file access, and no dangerous capabilities declared.
技能名称neural-memory
分析耗时23.3s
引擎pi
✓
可以安装
This skill is safe to use. The three sub-skills (memory-audit, memory-evolution, memory-intake) are well-documented workflow guides that only use custom NeuralMemory tools (nmem_*) with no shell, filesystem, or network access required.
资源类型声明权限推断权限状态证据
文件系统 NONE NONE — No filesystem access declared; ~/.neuralmemory/config.toml referenced only as co…
网络访问 NONE NONE — No network access tools declared; external URLs are documentation links only
命令执行 NONE NONE — No shell execution declared or implied
环境变量 NONE NONE — No environment variable access declared
技能调用 READ READ ✓ 一致 Uses nmem_* tools only via NeuralMemory MCP server
剪贴板 NONE NONE — No clipboard access declared
浏览器 NONE NONE — No browser access declared
数据库 NONE NONE — NeuralMemory internal graph database not exposed to skill
2 项发现
🔗
中危 外部 URL 外部 URL
https://nhadaututtheky.github.io/neural-memory
SKILL.md:40
🔗
中危 外部 URL 外部 URL
https://marketplace.visualstudio.com/items?itemName=neuralmem.neuralmemory
SKILL.md:41

目录结构

8 文件 · 31.4 KB · 935 行
Markdown 5f · 838L JSON 3f · 97L
├─ ▾ 📁 hooks
│ └─ 📋 hooks.json JSON 22L · 555 B
├─ ▾ 📁 skills
│ ├─ ▾ 📁 memory-audit
│ │ └─ 📝 SKILL.md Markdown 226L · 8.4 KB
│ ├─ ▾ 📁 memory-evolution
│ │ └─ 📝 SKILL.md Markdown 317L · 10.6 KB
│ ├─ ▾ 📁 memory-intake
│ │ └─ 📝 SKILL.md Markdown 179L · 6.1 KB
│ └─ 📝 README.md Markdown 75L · 1.8 KB
├─ 📋 marketplace.json JSON 43L · 1.2 KB
├─ 📋 plugin.json JSON 32L · 1.1 KB
└─ 📝 SKILL.md Markdown 41L · 1.6 KB

安全亮点

✓ No executable code - all files are pure documentation/metadata
✓ No shell, filesystem, or network access tools declared
✓ All allowed tools (nmem_*) are custom NeuralMemory operations that must be provided by an MCP server
✓ No credential harvesting, data exfiltration, or remote code execution patterns
✓ Skills use structured memory types and provide clear audit/pruning/creation workflows
✓ No base64, no eval(), no curl|bash, no subprocess patterns
✓ Context file (~/.neuralmemory/config.toml) is read-only reference, not access target
✓ MIT licensed with transparent repository references