ClawHub
Back to skill

Security audit

Structured PR Review

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed GitHub PR review helper that can read PR data and, when asked to address feedback, make normal PR updates such as commits, replies, thread resolution, and pushes.

Install this only if you want an agent to help both review PRs and actively address review feedback. Before using addressing mode, confirm the target repository, branch, generated code changes, comments, thread resolutions, and PR description edits before pushing or posting through gh.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
Findings (2)

Description-Behavior Mismatch

Medium
Confidence
92% confidence
Finding
The workflow is framed as responding to review comments, but it instructs the agent to modify code, create commits, push branch updates, resolve threads, and update PR metadata. Those are state-changing actions on both the local repository and remote hosting platform, so a review-oriented skill could be used to drive unauthorized or insufficiently reviewed code and PR changes beyond comment handling.

Context-Inappropriate Capability

Medium
Confidence
89% confidence
Finding
The documented steps grant repository-writing and remote mutation capability that is not tightly scoped to the stated purpose of structured PR review. If the agent follows these instructions automatically, it could alter source code, publish commits, post replies, resolve review threads, and change PR state based on untrusted review content or ambiguous user intent.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.