ClawHub
Back to skill

Security audit

微排版

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed workflow for creating and updating Weipaiban works, with sensitive API use that fits its stated purpose.

Install only if you are comfortable giving it a Weipaiban API key and, if using image generation, Volcengine/Jimeng credentials. Review generated text, images, and the final change summary before approval, and delete /tmp/weipaiban-task-* or ~/.u2net/u2netp.onnx afterward if the project contains sensitive content.

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (2)

Missing User Warnings

Medium
Confidence
92% confidence
Finding
The workflow explicitly sends generated or processed image/content data to external services (即梦 API, rembg processing flow, and a CDN) but does not mention any user notice, consent, or data handling limitations. In a content-authoring skill, users may provide copyrighted, sensitive, or personal material, so silent transfer to third parties creates privacy, compliance, and trust risks even if the behavior is part of normal functionality.

External Transmission

Medium
Category
Data Exfiltration
Content
```bash
# 受控执行模板：仅提交白名单字段
curl -s -X POST \
  -H "Authorization: Bearer $WEIPAIBAN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
Confidence
91% confidence
Finding
The step instructs the agent to perform an authenticated remote write using a bearer token and to transmit assembled content to an external API endpoint. Even though the document adds guardrails like user confirmation and a field whitelist, this still creates a real capability for outbound data transmission and state-changing API calls; if earlier task artifacts are poisoned or insufficiently validated, the agent could push unauthorized or attacker-controlled content to a live work item.

Static analysis

No suspicious patterns detected.