ClawHub
Back to skill

Security audit

Office Hours

Security checks across malware telemetry and agentic risk

Overview

This is a startup diagnostic skill that performs disclosed web research, limited project-context reading, and a bounded report write, with no evidence of hidden or malicious behavior.

Install this only where it is acceptable for an agent to read product docs and recent git history, perform web searches about the idea, and save a diagnostic report in the workspace. For confidential ideas or private repositories, ask the agent to skip web search, avoid git inspection, or return the report inline instead of writing the file.

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 (3)

Description-Behavior Mismatch

Medium
Confidence
94% confidence
Finding
The skill says it produces a diagnostic report, but it also instructs the agent to perform web searches and inspect local project files. That expands the data-access and action scope beyond what a user may reasonably expect from a startup-idea diagnostic, creating risk of unintended disclosure of repository contents or unnecessary external requests.

Context-Inappropriate Capability

Medium
Confidence
90% confidence
Finding
Instructing the agent to read project files and run `git log --oneline -20` introduces local inspection and command execution that are not essential to every office-hours conversation. Even though the command is read-only, it can expose sensitive metadata such as project names, commit subjects, internal roadmap details, or confidential initiatives.

Missing User Warnings

Medium
Confidence
92% confidence
Finding
The skill instructs saving output to `docs/business/office-hours-report.md` without first warning the user or obtaining consent. Any automatic write to the workspace can create unwanted files, overwrite prior work, leak sensitive discussion content into version-controlled directories, or trigger downstream tooling that watches the repository.

VirusTotal

63/63 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.