Proof of Work

• Outcome: Pass
• Latest audit: May 29, 2026, 8:32 PM
• Version:
  1.0.0

Overview

This is a local verification helper with an optional Ollama quality check; the main caveat is privacy clarity, not hidden or malicious behavior.

Install from the bundled files, not a remote curl-to-bash snippet. Keep check paths limited to agent output folders, review logs if outputs are sensitive, and only use --ai-check on files you are comfortable sending to your configured Ollama/model runtime.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (3)

Context-Inappropriate Capability

Medium

• Confidence: 94% confidence
• Finding: The optional AI quality check reads up to 2000 bytes from the target file and sends that content to `ollama run llama2` for analysis. Even if Ollama is local, this is a data-sharing boundary beyond simple validation, and the script does not constrain what kinds of files may be analyzed, so sensitive contents could be exposed to another service or model runtime.

Missing User Warnings

Medium

• Confidence: 85% confidence
• Finding: The README advertises AI-powered quality checks via Ollama/Heartbeat Kit but does not clearly warn that checked file contents may be transmitted to another local or remote processing component. Users may enable the feature on sensitive agent outputs without understanding the privacy and data-handling implications, creating an avoidable disclosure risk.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The script's `--ai-check` path sends file content to Ollama but the implementation itself does not present a strong user-facing disclosure at the point of use about content being shared with an AI service. In an agent skill context, that matters more because users may run the tool on arbitrary workspace outputs that can contain credentials, proprietary data, or personal information.

VirusTotal

65/65 vendors flagged this skill as clean.