ClawHub
Back to skill

Security audit

Safe Action

Security checks for vulnerabilities and agentic risk

Overview

This skill provides an advisory safety check before risky actions and clearly discloses its third-party API calls without bundling executable code.

Before installing, be comfortable with AgentUtil receiving action descriptions, platform/resource names, and timezone metadata, and avoid putting secrets, customer data, or sensitive internal details in the description field. Also note the disclosed per-check x402 pricing for paid service calls.

A.I.G

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

YARA rule 'agent_skill_mcp_tool_poisoning_metadata': MCP/tool metadata poisoning indicators in tool schemas or skill manifests [agent_skills]

High
Category
YARA Match
Content
---
name: safe-action
description: Before any destructive or irreversible action, run a safety pre-flight — check risks, reversibility, and timing.
version: 1.0.0
metadata:
  openclaw:
    emoji: "🛡️"
    homepage: https://agentutil.net
    always: false
---

# safe-action

Measure twice, cut once. Before taking destructive, irreversible, or high-stakes actions, this skill runs a three-part safety pre-flight: risk assessment, reversibility check, and timing awareness.

Combines three AgentUtil services — think (safety checklists), undo (reversibility intelligence), and context (situational timing) — into
Confidence
80% confidence
Finding
YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).

External Transmission

Medium
Category
Data Exfiltration
Content
**HTTP fallback:**
```bash
curl -X POST https://think.agentutil.net/v1/check \
  -H "Content-Type: application/json" \
  -d '{"action": "database_migration", "description": "Drop column users.email in production"}'
```
Confidence
94% confidence
Finding
The skill instructs the agent to send action metadata and free-form action descriptions to external HTTP endpoints. Even though the skill claims no user-generated content or credentials are transmitted, the example includes a production-specific schema change description, and in practice operators may include sensitive business, infrastructure, or security details in these fields. This creates a real data exfiltration and third-party dependency risk.

Static analysis

No suspicious patterns detected.