ClawHub
Back to skill

Security audit

eval-integrity

Security checks across malware telemetry and agentic risk

Overview

This is a read-only benchmark auditing skill with a disclosed GitHub PR lookup that is purpose-aligned but worth users understanding before use.

Installers should expect the skill to inspect the benchmark repository and, when GitHub CLI is configured, query open PRs for the target repo. It should not edit files or publish results during an audit; approve any proposed fixes separately.

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
93% confidence
Finding
The manifest and invocation text frame the skill as a local benchmark-repo audit that greps the target repo and reports findings. However, Step 1 requires running `gh pr list`, which reaches out to GitHub and inspects remote repository state beyond the local repo contents.

Context-Inappropriate Capability

Low
Confidence
82% confidence
Finding
The core purpose described is to audit methodology and integrity practices in a benchmark repository using repository evidence. Requiring `gh pr list` and verifying in-flight PR diffs adds a remote project-management/review capability that is not obviously necessary to determine current benchmark integrity, especially when the audit is described primarily as a repo grep/static-analysis workflow.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.