← 返回 Skills 市场
101
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install skillforge-meta
功能描述
Stop repeating yourself. SkillForge watches how you work, discovers your patterns, and forges them into reusable Skills — automatically. The more you use it,...
安全使用建议
What to check before enabling SkillForge:
- Confirm the runtime will only run the described read operations and will not transmit your files to external endpoints; this skill is instruction-only and the files included make no network calls, but the agent running it could be extended to do so. Ask: where does my agent run, and is network access restricted?
- Inspect the directories it will read: {workspace}/.workbuddy/memory/, {workspace}/.workbuddy/skillforge/, and ~/.workbuddy/skills/. Remove or redact any secrets or sensitive content from those logs if you don’t want them scanned.
- Verify the drafts directory (skillforge-drafts/) is indeed isolated and that generated drafts will require your explicit confirmation before installation (the SKILL.md promises this, but confirm your agent enforces it).
- Decide whether you want periodic/automatic scans. The config references auto_scan_enabled: true — if you prefer manual scans only, make sure to disable auto-scan in the config or agent settings.
- Monitor newly created files (drafts, patterns, skill-health.json) after the first run so you can confirm behavior matches documentation.
If you want higher assurance, request a code-based implementation (not instruction-only) or run the skill in a restricted/testing workspace first. If you are uncomfortable with the agent reading your workspace or home skills, do not enable it.
功能分析
Type: OpenClaw Skill
Name: skillforge-meta
Version: 1.0.0
The skillforge-meta bundle implements a 'self-evolving' agent system designed to monitor user activity logs and automatically generate new skill bundles. While the documentation in SKILL.md and ALGORITHM.md outlines safety constraints like draft isolation and mandatory user confirmation, the skill possesses high-risk capabilities including broad read access to user memory (~/.workbuddy/memory/), automated generation of executable instructions (SKILL.md), and the setup of recurring scheduled scans. These features represent a significant attack surface for persistence and automated code injection within the agent environment, even though the stated intent is productivity.
能力评估
Purpose & Capability
SkillForge's claimed goal (discover repeated workflows from your logs and generate Skill drafts) aligns with the instructions and reference docs: it explicitly scans daily logs, long-term memory, existing Skills, clusters patterns, scores them, and writes draft SKILL.md files. There are no unrelated credentials or external services requested.
Instruction Scope
The SKILL.md and ALGORITHM.md explicitly instruct the agent to read files under {workspace}/.workbuddy/memory/*.md, {workspace}/.workbuddy/memory/MEMORY.md, and ~/.workbuddy/skills/*/SKILL.md and to use the current conversation summary (cb_summary). It also writes drafts to {workspace}/skillforge-drafts/ and stores patterns/reports under {workspace}/.workbuddy/skillforge/. These file reads/writes are coherent with the stated purpose, but the skill metadata declared no required config paths — the actual runtime instructions reference many paths the agent will access, which is a transparency mismatch you should be aware of.
Install Mechanism
This is an instruction-only skill with no install spec and no code files, so nothing new is written to disk by an installer. The risk surface is limited to what the agent is instructed to read/write in your workspace/home directory.
Credentials
No environment variables or external credentials are requested. The only sensitive access is to your workspace/home files (logs, memory, existing Skills). That access is proportionate to the purpose of pattern discovery, but those files can contain secrets or private content — the ability to read them is the primary sensitivity here.
Persistence & Privilege
always is false and the skill states it will not auto-install or delete Skills. It does write drafts and stores patterns/health under the workspace/.workbuddy/skillforge directory. The CONFIG_FULL reference also includes general.auto_scan_enabled: true and auto_scan_rrule defaults — this implies the design supports periodic scans; confirm whether your agent runtime will run those scans automatically and whether you want that behavior.
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 skillforge-meta - 安装完成后，直接呼叫该 Skill 的名称或使用
/skillforge-meta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
v1.0.0
Initial release: Scout pattern discovery, Smith skill forging, Sensei health monitoring
元数据
常见问题
SkillForge — Self-Evolving AI Skills 是什么？
Stop repeating yourself. SkillForge watches how you work, discovers your patterns, and forges them into reusable Skills — automatically. The more you use it,... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 101 次。
如何安装 SkillForge — Self-Evolving AI Skills？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install skillforge-meta」即可一键安装，无需额外配置。
SkillForge — Self-Evolving AI Skills 是免费的吗？
是的，SkillForge — Self-Evolving AI Skills 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
SkillForge — Self-Evolving AI Skills 支持哪些平台？
SkillForge — Self-Evolving AI Skills 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 SkillForge — Self-Evolving AI Skills？
由 JFrankLee（@jfranklee）开发并维护，当前版本 v1.0.0。
推荐 Skills