扫描报告
20 /100
scienceclaw-watch
Run a live multi-agent scientific collaboration session and return a full summary when complete
Documentation-only skill with no implementation code present. References ANTHROPIC_API_KEY but declares no sensitive capabilities; actual behavior depends on unverifiable external scripts.
可以安装
Provide the actual bin/scienceclaw-watch Python script for full code review. Without it, the skill cannot be validated against its documentation.
安全发现 3 项
| 严重性 | 安全发现 | 位置 |
|---|---|---|
| 低危 | No implementation code to audit 文档欺骗 | SKILL.md:1 |
| 低危 | Hardcoded path dependency with no existence checks 权限提升 | SKILL.md:31 |
| 提示 | Undocumented credential dependency 敏感访问 | SKILL.md:2 |
| 资源类型 | 声明权限 | 推断权限 | 状态 | 证据 |
|---|---|---|---|---|
| 文件系统 | NONE | READ+WRITE | ✓ 一致 | SKILL.md writes to $OUTPUT_DIR and reads session_summary.json |
| 命令执行 | NONE | WRITE | ✓ 一致 | SKILL.md executes bash commands with python3 and activates .venv |
| 网络访问 | NONE | UNKNOWN | ✓ 一致 | Uses ANTHROPIC_API_KEY suggesting API calls, but not documented |
| 环境变量 | NONE | READ | ✓ 一致 | References ANTHROPIC_API_KEY and SCIENCECLAW_DIR environment variables |
| 技能调用 | NONE | READ | ✓ 一致 | Mentions scienceclaw-post and scienceclaw-investigate as follow-up skills |
目录结构
1 文件 · 4.6 KB · 123 行 Markdown 1f · 123L
└─
SKILL.md
Markdown
安全亮点
✓ No direct shell command injection vectors observed in documentation
✓ No base64-encoded or obfuscated commands in documentation
✓ No credential harvesting or exfiltration patterns described
✓ No network IOCs or external IP addresses mentioned
✓ Output directory is sandboxed under run_exports/ with timestamps