← 返回 Skills 市场
Hermes Learning
作者
liuboacean
· GitHub ↗
· v4.1.0
· MIT-0
132
总下载
0
收藏
0
当前安装
2
版本数
在 OpenClaw 中安装
/install hermes-learning
功能描述
Hermes 学习材料同步技能。从 Hermes Agent 获取自我更新后的学习材料，帮助 WorkBuddy 进行自我优化。支持 evolution.db 持久化、概念关联、双向反馈闭环。
使用说明 (SKILL.md)
Hermes 学习材料同步技能
此技能从 Hermes Agent 同步学习材料，帮助 WorkBuddy 进行自我更新和优化。
版本历史
- v4.1 (2026-04-20): 同步 Hermes 生成器 v3.6（评分 9.0+）；悬空助词 100% 消除；词汇边界截断优化；rule_id 唯一性机制完善；专有名词规范化（Hermes Agent / WorkBuddy Agent）
- v4 (2026-04-18): 修复 evolution.db 路径（全局路径）、移除无效 batch_size 自增、完善反馈数据结构
- v3: 双向反馈闭环、evolution.db 集成、概念关联
- v2: 策略库 + 效果追踪
- v1: 基础学习材料同步
学习材料来源
所有材料来自 Hermes 的记忆处理系统：
- 位置：
~/.hermes/shared/memory_summary.json - 反馈文件：
~/.hermes/shared/workbuddy_feedback.json - 策略库：
~/.workbuddy/skills/hermes-learning/strategies.json
evolution.db
- 路径:
~/.workbuddy/memory/evolution.db（全局统一路径） - 表: memories（学习条目）、concept_links（概念关联）
- 来源标识: source = 'hermes_learning'
使用方法
# 查看状态
python3 ~/.workbuddy/skills/hermes-learning/apply_learning.py status
# 应用学习材料
python3 ~/.workbuddy/skills/hermes-learning/apply_learning.py apply
# 列出所有策略
python3 ~/.workbuddy/skills/hermes-learning/apply_learning.py list
# 检查内容是否命中避免模式
python3 ~/.workbuddy/skills/hermes-learning/apply_learning.py check "要检查的内容"
# 查看反馈分析
python3 ~/.workbuddy/skills/hermes-learning/apply_learning.py feedback
双向反馈
- Hermes → WorkBuddy: 学习材料通过
memory_summary.json同步 - WorkBuddy → Hermes: 反馈通过
workbuddy_feedback.json回传 - 效果报告:
learning_effect_report.json记录应用效果
安全使用建议
This skill appears to do what it says: read Hermes-produced JSON summaries/feedback and import them as memories and concept links into a global evolution.db used by WorkBuddy. Before installing consider: 1) Back up ~/.workbuddy/memory/evolution.db and any WorkBuddy memory files you care about (the skill will create/modify that DB). 2) Inspect the full apply_learning.py (the prompt truncated the tail) yourself to confirm there are no network calls or unexpected behavior you missed. 3) Check the contents of ~/.hermes/shared/memory_summary.json and any feedback files for personally identifiable or sensitive data — the skill will ingest whatever is present. 4) If you are concerned about automatic changes to agent memory, avoid enabling autonomous/always-on invocation or run the script manually (the SKILL.md shows CLI commands: status/list/apply/check/feedback). 5) Note the minor documentation/code mismatch: changelogs claim env-var-based dynamic path discovery, but this version appears to use fixed home-directory paths; if you need different locations, review/modify the code before use.
功能分析
Type: OpenClaw Skill
Name: hermes-learning
Version: 4.1.0
The hermes-learning skill is designed to synchronize 'learning materials' and 'strategies' between two AI agents (Hermes and WorkBuddy). The core logic in apply_learning.py manages a local SQLite database (evolution.db) and processes JSON-based insights to optimize agent behavior. The script operates strictly within the user's home directory (~/.hermes and ~/.workbuddy), lacks network capabilities, and contains no evidence of malicious execution, data exfiltration, or harmful prompt injection.
能力标签
能力评估
Purpose & Capability
The name/description promise (sync Hermes learning into WorkBuddy, evolution.db persistence, concept links, feedback loop) matches what the code and SKILL.md do: reading Hermes memory_summary and feedback files, updating strategies/effects files, and writing to a global evolution.db under ~/.workbuddy/memory. There are no unrelated credentials or external services requested.
Instruction Scope
Runtime instructions and code operate on local files under the user's home: ~/.hermes/shared/* and ~/.workbuddy/* and on files in the skill directory (strategies.json, learning_effects.json). This is coherent with the stated purpose. Note: the SKILL.md and included data mention support for environment-variable-driven path discovery (HERMES_HOME, WORKBUDDY_HOME, etc.) in some changelogs, but the visible code uses fixed Path.home()-based locations rather than reading those env vars — a documentation/code mismatch to be aware of. Behavior is file I/O only; no network endpoints or credential access observed in the provided content.
Install Mechanism
No install spec or remote downloads are present; the skill is instruction-only plus bundled local code files. Nothing is fetched from external URLs and no archives are extracted.
Credentials
The skill declares no required environment variables or credentials, which aligns with the code. However, SKILL.md changelogs reference environment-variable options for discovery in other versions; the current code does not require or read credentials. The skill does create and write a global DB (~/.workbuddy/memory/evolution.db), which is reasonable for the described purpose but is a privileged local artifact (see persistence_privilege).
Persistence & Privilege
The skill will create/update a global evolution.db under ~/.workbuddy/memory and write files under the skill directory and ~/.workbuddy. always:false (not force-included). Autonomous invocation is allowed by default — this is normal — so if enabled the skill could be triggered to modify the global evolution.db without additional prompts. This is expected for a memory-sync skill but is impactful: it can change agent memory/state.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install hermes-learning - 安装完成后，直接呼叫该 Skill 的名称或使用
/hermes-learning触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v4.1.0
- 同步 Hermes 生成器 v3.6（评分 9.0+）
- 完全消除悬空助词，优化表达；
- 词汇边界截断逻辑改进，提升语义准确性；
- 优化 rule_id 唯一性判定与管理机制；
- 统一专有名词规范（Hermes Agent / WorkBuddy Agent）
v4.0.0
Hermes 学习材料同步技能 v4.0.0 包含重要修复与优化：
- 修复 evolution.db 路径为全局统一路径，避免多实例混乱
- 移除无效的 batch_size 自增逻辑，提升同步效率
- 完善反馈数据结构，支持更丰富的效果分析与闭环
- 继续支持学习材料同步、概念关联及双向反馈
元数据
常见问题
Hermes Learning 是什么？
Hermes 学习材料同步技能。从 Hermes Agent 获取自我更新后的学习材料，帮助 WorkBuddy 进行自我优化。支持 evolution.db 持久化、概念关联、双向反馈闭环。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 132 次。
如何安装 Hermes Learning？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install hermes-learning」即可一键安装，无需额外配置。
Hermes Learning 是免费的吗？
是的，Hermes Learning 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Hermes Learning 支持哪些平台？
Hermes Learning 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 Hermes Learning？
由 liuboacean（@liuboacean）开发并维护，当前版本 v4.1.0。
推荐 Skills