ClawHub
Back to skill

Security audit

深度研究专业版

Security checks for vulnerabilities and agentic risk

Overview

This research skill is broadly useful on paper, but its generic triggers and under-scoped scheduled/API/knowledge-base features need user review before installation.

Install only if you want a broad enterprise research workflow. Before using it, require explicit approval for exec commands, pip installs, API-key use, scheduled jobs, and any knowledge-base indexing or export. Keep the REST API bound to localhost with access controls, avoid sending sensitive research content unless intended, and verify any referenced script yourself because no implementation script was included in the artifact.

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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)

Vague Triggers

High
Confidence
79% confidence
Finding
The trigger condition is so broad that an agent may invoke this skill for generic data analysis, reporting, or visualization tasks outside its stated deep-research purpose. In a skill that advertises exec, API integration, file handling, and network access, over-broad routing increases the chance of unnecessary command execution, unintended data access, or external transmission in the wrong context.

External Transmission

Medium
Category
Data Exfiltration
Content
python （请参考skill目录中的脚本文件） --port 8000
# ...
# 提交研究任务
curl -X POST http://localhost:8000/research \
  -H "Content-Type: application/json" \
  -d '{"topic": "市场研究", "depth": "thorough"}'
# ...
Confidence
84% confidence
Finding
The skill explicitly documents a workflow that submits research content over HTTP to a local REST API endpoint. Even though the example targets localhost, it normalizes external transmission of potentially sensitive research data and uses plaintext HTTP, which can become dangerous if the service is rebound, proxied, containerized, or exposed beyond the local host.

Static analysis

No suspicious patterns detected.