ClawHub
Back to skill

Security audit

Code Test Check

Security checks for vulnerabilities and agentic risk

Overview

This skill is a source-code review helper that reads user-selected project files and creates a Markdown verification report, with no evidence of hidden, destructive, or unrelated behavior.

Install if you want an agent to inspect local PRD/test files and selected source directories to produce a Markdown verification report. Before running it, confirm the exact directories to search and where the report should be written, especially if you do not want files created under docs or beside the requirement document.

SkillSpector

Vulnerability Patterns
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Description-Behavior Mismatch

Medium
Confidence
95% confidence
Finding
The skill promises '只读分析' and '绝不修改源码', but later instructs the agent to save a Markdown report to disk. This is a real integrity/safety issue because it creates a misleading trust boundary: users may approve the skill believing no filesystem writes will occur, when it actually performs a write operation.

Intent-Code Divergence

Medium
Confidence
96% confidence
Finding
This is the same underlying issue expressed as a policy contradiction: the skill says it is read-only and will not modify files, yet the workflow includes persisting a report. Even if the write is limited to a new report file, contradictory instructions can cause unsafe tool behavior and prevent informed user consent.

Missing User Warnings

Medium
Confidence
93% confidence
Finding
The skill instructs the agent to save a generated report file but does not prominently disclose this write side effect in the top-level description. Hidden or under-disclosed file creation is risky because it can surprise users, overwrite files if naming collides, or violate expectations in constrained environments.

Static analysis

No suspicious patterns detected.