可信 — 风险评分 5/100
上次扫描:1 天前 重新扫描
5 /100
self-improvement
Captures learnings, errors, and corrections to enable continuous improvement
A pure-documentation self-improvement skill that logs learnings to markdown files with explicit secret-avoidance policies and no code execution. No malicious behavior found.
技能名称self-improvement
分析耗时28.5s
引擎pi
可以安装
Approve for use. No action required.
资源类型声明权限推断权限状态证据
文件系统 WRITE WRITE ✓ 一致 SKILL.md:creates .learnings/ dir and markdown files in workspace root
命令执行 WRITE WRITE ✓ 一致 SKILL.md: mkdir -p and file creation commands — scoped to project .learnings/ on…
网络访问 NONE NONE No network tool usage — external URLs are documentation links only
环境变量 NONE NONE SKILL.md explicitly instructs: 'Never log secrets, tokens, private keys, environ…
剪贴板 NONE NONE No clipboard access declared or observed
浏览器 NONE NONE No browser tool usage
数据库 NONE NONE No database access
技能调用 NONE NONE No skill self-invocation
1 项发现
🔗
中危 外部 URL 外部 URL
https://agentskills.io/specification
SKILL.md:581

目录结构

1 文件 · 19.0 KB · 603 行
Markdown 1f · 603L
└─ 📝 SKILL.md Markdown 603L · 19.0 KB

安全亮点

✓ Explicitly prohibits logging secrets, tokens, private keys, environment variables, and full source/config files
✓ Single markdown file with no executable code, scripts, or binaries
✓ File writes are scoped to project-local .learnings/ directory only
✓ External URLs are read-only documentation links, not fetches or downloads
✓ No obfuscation, base64, eval, or hidden functionality
✓ No credential harvesting, data exfiltration, or privilege escalation
✓ Hook integration is fully opt-in and requires explicit user configuration
✓ git clone targets a named directory (~/.openclaw/skills/) with no auto-execution