Low Risk — Risk Score 20/100
Last scan:15 hr ago Rescan
20 /100
Self-Check
系统自检工具 - OpenClaw environment diagnostic tool
This is a legitimate system self-check tool for OpenClaw. It performs read-only diagnostics via subprocess but makes no unauthorized modifications or data exfiltration.
Skill NameSelf-Check
Duration44.1s
Enginepi
Safe to install
Document shell:WRITE permission requirement in SKILL.md. Consider replacing the curl|bash suggestion in fix output with safer manual instructions.

Findings 2 items

Severity Finding Location
Low
Shell permission not declared Doc Mismatch
SKILL.md lists node/npm/nvm binaries as required tools but does not explicitly declare shell:WRITE permission needed for subprocess execution
requires: bins:[node,npm,nvm]
→ Add 'allowed-tools: Bash→shell:WRITE' to SKILL.md metadata
SKILL.md:1
Low
API key presence check via env iteration Sensitive Access
Script checks specific named API keys (OPENAI_API_KEY, etc.) in environment variables but does not read or exfiltrate values
env_keys = ['OPENAI_API_KEY', 'MINIMAX_API_KEY', ...]
→ Document this behavior in SKILL.md under API Token section
scripts/self_check.py:265
ResourceDeclaredInferredStatusEvidence
Shell NONE WRITE ✗ Violation scripts/self_check.py:78 - subprocess.run() for node/npm/pgrep checks
Filesystem READ READ ✓ Aligned Scripts only read configs at known paths
Environment NONE READ ✓ Aligned scripts/self_check.py:265 - checks os.environ for API key names
1 Critical 1 findings
💀
Critical Dangerous Command 危险 Shell 命令
curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.39.0/install.sh | bash
scripts/self_check.py:126

File Tree

2 files · 19.8 KB · 596 lines
Python 1f · 497L Markdown 1f · 99L
├─ 📁 scripts
│ └─ 🐍 self_check.py Python 497L · 17.2 KB
└─ 📝 SKILL.md Markdown 99L · 2.5 KB

Security Positives

✓ Core principle correctly implemented: only checks/reports/suggests, never auto-fixes
✓ No credential exfiltration - API keys checked only for presence, values never read or transmitted
✓ No network requests to external IPs from the script itself
✓ No base64, eval, obfuscation, or anti-analysis techniques
✓ Remote curl|bash is only in fix suggestion text, not executed
✓ All subprocess commands are benign diagnostic commands (node --version, pgrep, ss, etc.)
✓ No access to sensitive paths like ~/.ssh, ~/.aws, or .env files
✓ SKILL.md core principle statement is honest and accurately describes behavior