中文 EN

Scan Report

Scan New Files

⚠ Low Risk — Risk Score 15/100
Last scan:2 days ago Rescan
15 /100
proactive-companion
An autonomous, self-improving buddy system for OpenClaw that schedules interest-based Telegram pings
A legitimate proactive companion skill that schedules Telegram pings via OpenClaw cron, with no malicious patterns found. Minor concern about invasive SOUL.md patching during install.
Skill Nameproactive-companion
Duration43.3s
Enginepi
✓
Safe to install
Review the automatic SOUL.md modification during install. Ensure you are comfortable with the skill patching system files. Otherwise safe to use.

Findings 3 items

Severity Finding Location
Medium
Automatic SOUL.md Modification
The install.sh script automatically appends routing rules to /data/.openclaw/SOUL.md without explicit user consent beyond the install process.
cat >> "$SOUL_FILE" << 'SOULEOF' ...
→ Document this clearly during installation or provide a manual option.
 install.sh:148
Low
tools.profile = coding Modification
Install script attempts to change OpenClaw tools profile to 'coding' which enables broader execution capabilities.
$OPENCLAW_BIN config set tools.profile coding
→ This is necessary for the skill to function but increases attack surface.
 install.sh:105
Low
Broad subprocess Usage
The code uses subprocess to call openclaw CLI for session management, history retrieval, and trigger injection.
subprocess.Popen(["openclaw", "agent", ...])
→ While documented, subprocess usage is extensive. Ensure openclaw CLI is trusted.
 proaktiv_check.py:175
ResourceDeclaredInferredStatusEvidence
Shell NONE WRITE ✓ Aligned proaktiv_check.py:175-183 uses subprocess.Popen for openclaw agent command
Filesystem WRITE WRITE ✓ Aligned SKILL.md declares filesystem:/data/.openclaw/skills/proaktiv/ - code operates wi…
Network NONE NONE — No direct network calls; all Telegram communication via openclaw CLI
Environment READ READ ✓ Aligned Reads OPENCLAW_TELEGRAM_NR from .env file
Database NONE NONE — Uses JSON files for state storage within skill directory
2 findings
🔗
Medium External URL 外部 URL
https://clawhub.ai
.clawhub/origin.json:3
🔗
Medium External URL 外部 URL
https://t.me/userinfobot
SKILL.md:160

File Tree

16 files · 82.2 KB · 2085 lines
Python 4f · 1229L Markdown 4f · 559L Shell 1f · 214L JSON 6f · 75L Ignore 1f · 8L
├─ ▾ 📁 .clawhub
│ └─ 📋 origin.json JSON 7L · 141 B
├─ ▾ 📁 templates
│ ├─ 📋 interest_graph.json JSON 14L · 308 B
│ └─ 📋 social_knowledge.json JSON 3L · 19 B
├─ 📋 _meta.json JSON 6L · 148 B
├─ 📄 .gitignore Ignore 8L · 98 B
├─ 🐍 feedback_update.py Python 256L · 10.5 KB
├─ 🔧 install.sh Shell 214L · 8.3 KB
├─ 🐍 interest_evolve.py Python 114L · 3.9 KB
├─ 🐍 proaktiv_check.py Python 806L · 31.3 KB
├─ 🐍 proaktiv_onboarding.py Python 53L · 2.1 KB
├─ 📝 README.md Markdown 154L · 5.7 KB
├─ 📋 skill.json JSON 42L · 1.1 KB
├─ 📝 SKILL.md Markdown 284L · 12.2 KB
├─ 📋 social_knowledge.json JSON 3L · 19 B
├─ 📝 SOCIAL.md Markdown 34L · 1.4 KB
└─ 📝 TOPIC_TEMPLATES.md Markdown 87L · 5.1 KB

Security Positives

✓ No credential harvesting or exfiltration detected
✓ No base64, eval, or obfuscated code found
✓ No access to ~/.ssh, ~/.aws, or other sensitive paths
✓ No direct IP network requests or C2 communication
✓ No curl|bash or wget|sh remote script execution
✓ All data operations confined to skill directory /data/.openclaw/skills/proaktiv/
✓ Comprehensive changelog showing active development and bug fixes
✓ Clean, commented code with no hidden functionality
✓ MIT license properly declared
✓ Social knowledge tracking is transparent about data collection