ClawHub
Back to skill

Security audit

smartclaws

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed SmartClaws setup guide for blockchain IoT agents, with high-impact controls that are mostly owner-confirmed and purpose-aligned.

Before using this skill for a live device, fill in AGENTS.md with concrete session authority and keep command/publish permissions limited to trusted sessions. Review the SmartClaws tool allowlist after setup, especially write tools for publish, disclose, register, roles, viewing keys, and backups.

SkillSpector

Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Vague Triggers

Medium
Confidence
85% confidence
Finding
The template explicitly tells the agent it may ask the user how it should behave, but it does not constrain when configuration changes are allowed or require that such requests come from an authorized session. In a bridge agent that can publish telemetry or apply device commands, this ambiguity can let an untrusted user socially steer operating policy, cadence, or safety thresholds and potentially influence future on-chain actions.

Vague Triggers

Medium
Confidence
92% confidence
Finding
The authority section leaves the allowlist blank and says the owner should define who may trigger manual publish or commands, but the template provides no secure default beyond a comment. In practice, deployed agents may be instantiated with incomplete policy, causing command or publish operations to be governed by vague session context and making unauthorized triggering more likely.

Static analysis

No suspicious patterns detected.