by 下载技能包
最后更新
2026/4/22
安全扫描
OpenClaw
可疑
medium confidenceThe skill's code and instructions broadly match its stated goal (auto-observing and improving other skills), but it asks for powerful filesystem access, can autonomously edit other skills, contains inconsistencies between its promise and behavior, and the included analysis script appears truncated/buggy — you should review and limit its privileges before installing.
评估建议
Before installing or enabling Skill Garden, consider the following:
- This skill needs read/write access to your ~/.openclaw/workspace/skills tree and will read logs and edit other skills' SKILL.md files. That is necessary for its purpose but is a powerful privilege — back up your skills directory first.
- SKILL.md promises 'user-in-control' but the decision rules auto-apply edits at high and mid confidence (≥90% and 70–89%). If you want manual approval for all edits, change the min-confidence...详细分析 ▾
ℹ 用途与能力
Skill Garden's name and description match its implementation: it reads per-skill usage logs, evaluates skills, generates proposals, and edits SKILL.md files. Requiring read/write access to ~/.openclaw/workspace/skills is coherent with the stated purpose. However, the skill claims to be 'user-in-control' and 'transparent' while also auto-applying changes in multiple confidence bands (see instruction_scope).
⚠ 指令范围
SKILL.md instructs the agent to passively log every skill invocation and to run weekly batch analysis that can auto-apply edits to other skills' SKILL.md. This gives it the ability to collect broad cross-skill telemetry and to modify other skills without a per-change explicit approval (the decision tree auto-applies changes at ≥90% and even 70–89% as [experimental]). That scope is functionally necessary for an auto-improver, but it is a high-risk capability (silent/automatic edits to other skills, broad read/write of skill directories).
✓ 安装机制
There is no network download or external installer; the skill is instruction-only with bundled scripts. That's lower risk than arbitrary remote downloads. Files are intended to be run from the user's workspace and operate on local files only.
ℹ 凭证需求
The skill requests no environment variables or external credentials, which is proportionate to its stated function. However, it implicitly requires filesystem access to ~/.openclaw/workspace/skills (read and write), which is not listed in a formal manifest. That filesystem access is necessary for the purpose but is high-privilege because it touches other skills' files and the grower's own memory/logs.
⚠ 持久化与权限
always:false (good) but the skill is allowed to invoke autonomously and — by design — can modify other skills' SKILL.md and maintain its own cron schedule and memory. Modifying other skills' code/content is central to its purpose but is also a significant privilege; the SKILL.md contains inconsistent statements about when changes 'ask first' vs auto-apply, increasing the risk of unexpected permanent changes.
安全有层次，运行前请审查代码。
运行时依赖
无特殊依赖
版本
latestv1.0.02026/4/22
Initial release: automatic skill improvement through passive usage observation, weekly batch analysis, and targeted SKILL.md edits. Your skills don't just exist — they grow.
● 可疑
安装命令
点击复制官方npx clawhub@latest install skill-garden
镜像加速npx clawhub@latest install skill-garden --registry https://cn.longxiaskill.com