扫描报告
15 /100
bw-openclaw-boost
OpenClaw效率提升工具包 - 成本追踪、记忆管理、压缩系统、权限控制
A legitimate read-only efficiency toolkit for OpenClaw; the only notable issue is a stale SKILL.md reference to a removed stream_exec.py, and permission_manager.py permits curl/wget in its safe-commands list but flags piped variants as dangerous.
可以安装
Update SKILL.md to remove the stream_exec.py reference (v1.0.7 removed it) and tighten the permission_manager.py safe-commands list to exclude curl/wget entirely. Otherwise this skill is safe to use.
安全发现 4 项
| 严重性 | 安全发现 | 位置 |
|---|---|---|
| 低危 | Stale SKILL.md reference to removed tool | SKILL.md:1 |
| 低危 | Broad safe-commands allowlist in permission_manager | tools/permission_manager.py:48 |
| 低危 | slash_commands.py references removed stream_exec.py | tools/slash_commands.py:93 |
| 提示 | subprocess usage not declared in SKILL.md | tools/token_budget.py:40 |
| 资源类型 | 声明权限 | 推断权限 | 状态 | 证据 |
|---|---|---|---|---|
| 文件系统 | NONE | WRITE | ✓ 一致 | memory_manager.py, memory_relevance.py, dream_consolidation.py all write to ~/.o… |
| 网络访问 | NONE | NONE | — | No network I/O detected in any tool. subprocess calls only invoke 'openclaw stat… |
| 命令执行 | NONE | READ | ✓ 一致 | install.sh and launch.sh execute but only for installation and tool dispatch; SK… |
| 环境变量 | NONE | NONE | — | No environment variable access observed |
| 技能调用 | NONE | NONE | — | No cross-skill invocation |
| 剪贴板 | NONE | NONE | — | No clipboard access |
| 浏览器 | NONE | NONE | — | No browser access |
| 数据库 | NONE | NONE | — | No database access |
目录结构
16 文件 · 111.2 KB · 3550 行 Python 11f · 3252L
Shell 3f · 135L
Markdown 1f · 128L
JSON 1f · 35L
├─
▾
tools
│ ├─
check_permission.sh
Shell
│ ├─
compaction_manager.py
Python
│ ├─
coordinator.py
Python
│ ├─
cost_tracker.py
Python
│ ├─
dream_consolidation.py
Python
│ ├─
feature_flags.py
Python
│ ├─
memory_manager.py
Python
│ ├─
memory_relevance.py
Python
│ ├─
permission_manager.py
Python
│ ├─
slash_commands.py
Python
│ ├─
token_budget.py
⚠
Python
│ └─
tool_tracker.py
Python
├─
install.sh
Shell
├─
launch.sh
Shell
├─
SKILL.md
Markdown
└─
version.json
JSON
安全亮点
✓ No network exfiltration or data theft behavior detected
✓ No credential harvesting (no access to ~/.ssh, ~/.aws, .env)
✓ No base64/eval obfuscation or obfuscated payloads
✓ No reverse shell or C2 infrastructure indicators
✓ No curl|bash or wget|sh remote script execution
✓ Dangerous functionality (stream_exec) was actively removed in recent version
✓ dream_consolidation defaults to dry-run mode and requires --force flag to delete
✓ Configuration files are scoped to the skill directory (~/.openclaw/bw-openclaw-boost/) not system-wide
✓ Permission manager correctly identifies and blocks fork bombs, rm -rf, dd block writes
✓ All openclaw CLI invocations are read-only (status, cron list) as documented