ClawHub

代码审查 89d9f5

Security checks across malware telemetry and agentic risk

Overview

This is a markdown-only code review learning guide with overly broad activation wording but no executable, credential, persistence, or data-access behavior.

Install only if you want a lightweight Chinese code review tutorial skill. Be aware it may activate on generic OpenAI or code-review conversations; narrowing the trigger to explicit Codex code-review tutorial wording would reduce noise.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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 (2)

Vague Triggers

High
Confidence
97% confidence
Finding
The top-level trigger includes very broad phrases such as '代码审查 / OpenAI / Codex教程', which are common conversational terms and likely to match unrelated user requests. This can cause unintended skill activation, injecting irrelevant or untrusted guidance into unrelated sessions and creating a prompt-scope hijack risk if the skill is auto-loaded by keyword.

Vague Triggers

Medium
Confidence
98% confidence
Finding
The activation conditions are ambiguous and overly permissive, including ordinary phrases like '用户说"代码审查"' and '用户说"OpenAI"'. In a system that auto-selects skills from markdown conditions, this increases the chance of accidental activation and cross-context prompt injection from a skill whose content is largely aggregated, duplicated, and not tightly curated.

VirusTotal

62/62 vendors flagged this skill as clean.

View on VirusTotal