HERMES - AI Agent 协作进化系统

HERMES is an AI agent platform enabling proactive skill prediction, dynamic composition, decentralized trust, and version-controlled collaborative evolution.

Security Scan

The package claims a complex, networked evolution platform but ships incomplete code and no install/dependency information, making its behavior unclear and potentially unsafe to run as-is.

Purpose & Capability

The skill describes a multi-module, decentralized trust platform but the repository only contains a thin CLI and index.js; required submodules (./hermes-core, ./hermes-predict, ./hermes-compose, ./hermes-version-tree, ./hermes-web-trust, etc.) are referenced but not present and package.json lists no dependencies. That mismatch means the shipped files cannot implement the claimed capabilities and is not consistent with the stated purpose.

Instruction Scope

SKILL.md and the CLI show commands and examples (predict, compose, leaderboard, install via 'clawhub install hermes-full'), but the runtime instructions rely on code that is missing or references paths that don't exist (e.g., 'node test/validate.js' in examples). The instructions do not request secrets or external endpoints, but they assume functionality that is not present, which is misleading and gives the agent broad, undefined discretion if the missing pieces are supplied at runtime.

Install Mechanism

There is no install specification despite a CLI and JS source files being present. package.json is minimal and declares no dependencies, yet index.js requires multiple internal modules that are absent. This inconsistency could indicate an incomplete package, or that additional code would be fetched/installed from elsewhere — neither of which is documented here.

Credentials

The skill declares no required environment variables or credentials, which superficially limits sensitive access. However, the described decentralized trust and web-trust components would normally require network access and possibly keys — their absence (and absence of module code) is inconsistent and raises questions about where those capabilities would be implemented and whether credentials would be requested later.

Persistence & Privilege

The skill does not request always:true or any elevated platform privileges and is user-invocable only. There is no evidence it modifies other skills or global agent settings in the provided files.

What to consider before installing

Do not install or run this package in a production environment yet. Key concerns: (1) multiple required submodules referenced by index.js are missing, so the package is incomplete or expects external code; (2) package.json contains no dependencies or install instructions — ask the author for the missing module files, a full dependency list, and an explicit install mechanism; (3) request source provenance: who published this, a homepage/repo, and cryptographic signing or checksums; (4) ask for explanations of network behavior (does the web-trust module contact external services?) and any credential requirements; (5) if you must test, do so in an isolated sandbox (offline VM or container) and inspect any additional modules before executing. If the author cannot provide the missing modules and a reproducible install method, treat the package as untrustworthy.

超越 EvoMap 的下一代 AI Agent 知识共享与进化平台

🎯 核心创新

特性	EvoMap	HERMES
触发方式	被动反馈	主动预测
技能结构	单一 Capsule	动态组合
版本控制	无	分支+合并
信任机制	中心化 Hub	去中心化

🏗️ 四层架构

Instance → Meta → Gene → Capsule

🚀 快速开始

# 安装
clawhub install hermes-full

# 预测任务
hermes predict "写一个安全的 API"

# 组合技能
hermes compose auth validation security

# 查看信任排行榜
hermes leaderboard

📦 包含模块

• hermes-core - 核心引擎
• hermes-predict - 预测学习
• hermes-compose - 技能组合
• hermes-version-tree - 版本控制
• hermes-web-trust - 信任网络