ClawHub
Back to skill

Security audit

JSON修复工具

Security checks for vulnerabilities and agentic risk

Overview

This JSON repair skill has a coherent purpose, but it asks for command and file-write authority while giving overbroad and partly unsafe guidance about recursive overwrites and JavaScript-based parsing.

Use this only on user-specified, trusted JSON-like files, preferably with backups and a small path scope. Avoid using it on files from untrusted sources unless the implementation uses a dedicated non-executing parser, and treat recursive repair or --no-backup as high-risk operations.

SkillSpector

Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • 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
Findings (3)

Intent-Code Divergence

Medium
Confidence
82% confidence
Finding
The skill claims comments are safely handled by the standard Node parser and elsewhere promotes parsing loose JSON via vm.runInNewContext. Misstating parser behavior around untrusted input can lead operators to process attacker-controlled files under unsafe assumptions, increasing the risk of executing JavaScript-like payloads or accepting formats that are not actually safe to parse this way.

Vague Triggers

Medium
Confidence
78% confidence
Finding
The skill is advertised with broad trigger language such as generic productivity, automation, and workflow optimization, while it includes exec, read, and write capabilities. In agent environments, this can cause over-selection of a file-modifying skill for loosely related prompts, raising the chance of unintended command execution or file changes on sensitive user data.

Missing User Warnings

Medium
Confidence
88% confidence
Finding
The documentation describes overwriting files and even disabling backups without a prominent warning, despite the skill having write and exec capabilities. In this context, understated disclosure of destructive behavior is dangerous because users or calling agents may invoke the tool without realizing it can recursively modify files or remove the recovery safeguard, leading to data loss or tampering.

Static analysis

No suspicious patterns detected.