中文 EN

Scan Report

Scan New Files

✓ Trusted — Risk Score 5/100
Last scan:18 hr ago Rescan
5 /100
ai-interview
AI面试系统 - 完整的AI面试解决方案，包含求职者和面试官Agent，支持飞书群聊面试和实时可视化观察
Legitimate AI interview system with job-seeker and recruiter agents, featuring a local web visualization panel. No malicious behavior, credential harvesting, or external network communication detected.
Skill Nameai-interview
Duration45.9s
Enginepi
✓
Safe to install
This skill is safe to use. The web viewer server binds to localhost only (port 8091) and accesses only the intended OpenClaw session directory.

Findings 2 items

Severity Finding Location
Low
Reads OpenClaw Agent Session Files Sensitive Access
The server reads JSONL session files from ~/.openclaw/agents/<agent>/sessions/ to display conversation history. These files may contain sensitive conversation content.
agent_dir = SESSION_DIR / agent_name / "sessions"
→ This is necessary for the visualization feature. Consider documenting that session files contain conversation history.
 server.py:53
Low
Can Delete Session Files Sensitive Access
The /api/clear endpoint can delete agent session files via clear_conversations() function.
s.unlink()
→ User-initiated clear functionality is acceptable. No automatic deletion observed.
 server.py:80
ResourceDeclaredInferredStatusEvidence
Filesystem NONE READ ✓ Aligned server.py:26 reads from ~/.openclaw/agents
Network NONE READ ✓ Aligned server.py:16 localhost HTTP server only
Shell NONE NONE — install.sh:41 uses subprocess for background launch, documented

File Tree

6 files · 27.9 KB · 905 lines
HTML 1f · 325L Markdown 3f · 278L Python 1f · 246L Shell 1f · 56L
├─ ▾ 📁 config
│ ├─ ▾ 📁 job-seeker
│ │ └─ 📝 IDENTITY.md Markdown 23L · 516 B
│ └─ ▾ 📁 recruiter
│ └─ 📝 IDENTITY.md Markdown 31L · 612 B
├─ ▾ 📁 public
│ └─ 📄 index.html HTML 325L · 11.7 KB
├─ ▾ 📁 scripts
│ └─ 🔧 install.sh Shell 56L · 1.5 KB
├─ 🐍 server.py Python 246L · 8.2 KB
└─ 📝 SKILL.md Markdown 224L · 5.4 KB

Security Positives

✓ No external network communication - server only binds to localhost
✓ No credential harvesting or environment variable iteration for secrets
✓ No base64 encoding or obfuscation
✓ No reverse shell or C2 infrastructure
✓ No curl|bash or wget|sh remote execution patterns
✓ Clear documentation matches implementation
✓ No malicious dependencies or supply chain risks
✓ Session file access is scoped to OpenClaw directory only
✓ Install script only creates directories and copies templates