ClawHub
Back to skill

Security audit

分析

Security checks across malware telemetry and agentic risk

Overview

This skill claims to check GitHub PR mergeability, but it is too broadly scoped and asks for write and command execution authority without clear limits.

Review before installing. Use only in a constrained environment, and do not grant write or command execution authority unless the publisher narrows the skill to GitHub PR analysis, documents exact commands or scripts, and explains credential handling for private repositories.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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 (4)

Description-Behavior Mismatch

High
Confidence
95% confidence
Finding
The manifest presents the skill as a narrow GitHub PR mergeability tool, while the body expands it into generic data analysis, file processing, API use, and command execution. This scope deception is dangerous because agents or users may grant broader trust and permissions than warranted, enabling misuse under the guise of a harmless code-review utility.

Context-Inappropriate Capability

High
Confidence
97% confidence
Finding
Advertising arbitrary file writing and generic command execution for a PR mergeability skill violates least privilege and creates a clear pathway for unintended or abusive actions. In an agent environment, these capabilities could be used to modify local files, run unsafe shell commands, or pivot into broader system access unrelated to the stated task.

Vague Triggers

High
Confidence
93% confidence
Finding
The broad 'Use when' trigger is unrelated to PR mergeability and can cause the skill to be invoked in many contexts involving general analysis or reporting. Overbroad invocation increases the chance that an agent will route unrelated tasks to a skill that has elevated tools, expanding the attack surface and creating opportunities for unintended execution.

Vague Triggers

Medium
Confidence
91% confidence
Finding
The instructions describe broad automation actions like file writing, API integration, and command execution without clear limits or activation conditions. In agentic systems, vague capability descriptions can lead to excessive autonomy and unsafe task expansion beyond the user's intended PR-analysis request.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.