ClawHub
Back to skill

Security audit

WorkBuddy Usage Status

Security checks across malware telemetry and agentic risk

Overview

This skill locally summarizes WorkBuddy usage into an offline dashboard, with no artifact evidence of exfiltration or destructive behavior.

Install only if you want a local WorkBuddy usage dashboard. Run it deliberately, prefer an explicit output folder you control, and avoid sharing or syncing the generated report files unless you are comfortable exposing your usage history, session titles, model choices, costs, timestamps, and error summaries.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Intent-Code Divergence

Medium
Confidence
98% confidence
Finding
The template inserts summary fields into the DOM with innerHTML using unescaped values from window.USAGE_STATUS.summary, including generated_at, date_min, and date_max. If the upstream data generator ever includes attacker-controlled strings, this becomes a straightforward DOM XSS sink that executes when the local dashboard is opened, and the 'pure local' context can lower user suspicion and increase trust in the rendered content.

Vague Triggers

Medium
Confidence
93% confidence
Finding
The auto-detection trigger list includes very generic terms such as "telemetry" and especially "usage", which can match many ordinary conversations unrelated to this skill. In an agent environment, broad triggers can cause unintended activation and execution of the skill, leading to unnecessary access to local usage data and surprising behavior without clear user intent.

Vague Triggers

Medium
Confidence
94% confidence
Finding
The Chinese invocation examples repeat broad generic triggers, including "telemetry", "usage", and "思考效率", without constraints tying them specifically to WorkBuddy status reporting. This increases the chance that normal discussion about efficiency, telemetry, or usage will unintentionally invoke the skill and expose local analytics or consume resources parsing local trace data.

Missing User Warnings

Medium
Confidence
90% confidence
Finding
The script enumerates and aggregates local WorkBuddy session metadata and trace data from ~/.workbuddy, then writes the results to JSON/JS/HTML files in a user-chosen output directory. Even though it is local-only and does not exfiltrate data itself, the generated artifacts can expose sensitive usage patterns, model names, timestamps, session titles, errors, and token/credit history to other local users, shared folders, backups, or accidental publication, and there is no explicit warning, consent gate, or minimization control.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.