ClawHub
Back to skill

Security audit

psych-questionnaire-cleaner

Security checks for vulnerabilities and agentic risk

Overview

This is a documentation-only data-cleaning skill whose sensitive behavior is disclosed, bounded to user-provided questionnaire datasets, and designed to preserve raw data with audit records.

Install this skill when Chinese-language questionnaire cleaning outputs are appropriate for your workflow. Before use, confirm that downstream analysts can work with Chinese filenames and column names, and review the predefined response-time and attention-check exclusion rules because they can remove rows from the analytical dataset while preserving them in separate audit records.

SkillSpector

Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • 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)

Natural-Language Policy Violations

Medium
Confidence
91% confidence
Finding
The skill hard-requires Simplified Chinese for filenames, column names, logs, and reports without indicating user choice or consent. This can create accessibility, usability, and governance problems for users or downstream systems that expect another language, increasing the risk of misinterpretation of data-cleaning outputs and audit artifacts.

Natural-Language Policy Violations

Medium
Confidence
93% confidence
Finding
The template hard-codes `language: "zh-CN"` and `rename_output_columns_to_chinese: true` as output settings, which can override user expectations and cause silent localization of reports and column names. In a data-cleaning skill, forced locale/renaming can degrade auditability, break downstream pipelines, and increase the chance of misinterpretation or privacy/compliance mistakes when collaborators expect original-language outputs.

Static analysis

No suspicious patterns detected.