ClawHub
Back to skill

Security audit

workspace-architect

Security checks across malware telemetry and agentic risk

Overview

This skill is a workspace-configuration helper whose file access and sandboxed modification model are mostly disclosed and aligned with its purpose.

Before installing, be aware that meaningful workspace analysis may expose your OpenClaw configuration, profile, memory, and tool notes to the agent running the skill. Use it when you intend that review, check drafts before applying them, and keep secrets out of workspace files.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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
95% confidence
Finding
This is a significant trust-boundary mismatch: the skill promises sandbox-only analysis and guided creation/optimization, but the finding says it actually reads real workspace files from ~/.openclaw/workspace and performs much less analysis than claimed. Users may expose sensitive configuration, prompts, secrets, or personal data under false assumptions about isolation and scope, which is dangerous even without direct file modification.

Description-Behavior Mismatch

High
Confidence
97% confidence
Finding
The skill explicitly claims that all work occurs in a sandbox without touching originals, but the analyze command reads from the user's live ~/.openclaw/workspace directory. That creates a trust-boundary violation: users may expose real configuration contents under the assumption that only sandboxed data is used, which can lead to unintended disclosure of sensitive workspace information.

Intent-Code Divergence

High
Confidence
98% confidence
Finding
The documentation states all work is done in a sandbox, yet the code inspects the live workspace under the user's home directory. In a security-sensitive agent skill, misleading assurances about sandboxing are dangerous because users and higher-level tooling may authorize execution based on false assumptions about isolation and data exposure.

Vague Triggers

Medium
Confidence
80% confidence
Finding
The usage section suggests very broad natural-language triggers like 'Analyze my workspace configuration' and 'Optimize my workspace files' without defining a clear invocation boundary or requiring explicit activation context. In an agent ecosystem, this can cause the skill to engage during ordinary conversation or be invoked unintentionally by adjacent user text, increasing the chance of unintended analysis or workflow execution.

VirusTotal

61/61 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.