ClawHub
Back to skill

Security audit

Openclaw Cache Kit

Security checks across malware telemetry and agentic risk

Overview

The skill makes local OpenClaw caching changes that are mostly disclosed and purpose-aligned, but users should understand it edits global config, restarts the gateway, and enables local diagnostics.

Install only if you want a local script to change your global OpenClaw defaults. Before running apply.sh, review the config changes, keep the backup path, and check ~/.openclaw/logs/cache-trace.jsonl for sensitive operational details after diagnostics are enabled.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Tp4

High
Category
MCP Tool Poisoning
Confidence
84% confidence
Finding
The skill description understates operationally sensitive behavior: modifying local configuration, enabling detailed diagnostics logging, parsing local logs, and restarting the gateway. Hidden or insufficiently disclosed side effects increase the risk that users execute the skill without understanding it changes runtime behavior, expands local telemetry collection, or may affect availability.

Description-Behavior Mismatch

Medium
Confidence
86% confidence
Finding
The skill claims to apply caching optimizations, but it also changes model aliasing and enables diagnostics settings that go beyond a narrowly scoped cache tweak. This mismatch increases the risk of surprising behavior, policy drift, or unintended data exposure because users may run the script expecting only performance-related changes.

Context-Inappropriate Capability

Medium
Confidence
92% confidence
Finding
Enabling diagnostic cache tracing is not necessary to achieve caching optimization and can increase observability into prompt or message handling. Even though includePrompt and includeMessages are disabled, turning on diagnostics may still create additional sensitive metadata exposure or logging surface that the user did not intend to enable.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The script directly overwrites the user's OpenClaw configuration and attempts a gateway restart without an interactive confirmation step. In a skill context, this is dangerous because users may invoke it expecting an advisory or reversible change, but it performs immediate persistent modifications that can disrupt service or alter security-relevant settings.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.