NXT Pulse Agent

• Outcome: Pass
• Latest audit: May 29, 2026 at 8:41 PM UTC
• Version:
  0.5.0

Overview

This is a coherent local energy-management skill, but users should be careful about which personal files they allow it to read.

Install only if you want an agent to infer your energy or availability from conversation context. Keep CONTEXT_SOURCES limited to specific, low-sensitivity files unless you intentionally want journals, logs, or health-related notes used for this purpose. Review memory/pulse-state.json and memory/pulse-history.jsonl periodically, and use /pulse quiet or disable the skill if proactive nudges become unwanted.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• 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 (4)

Missing User Warnings

Medium

• Confidence: 90% confidence
• Finding: The README advertises semantic monitoring of user messages and configurable external context sources, but it does not clearly disclose what data is read, when it is processed, how much is retained, or whether the user explicitly consents. In an agent skill, this can lead to covert collection or overbroad analysis of sensitive personal content such as journals and activity logs, increasing privacy and misuse risk.

Natural-Language Policy Violations

Low

• Confidence: 79% confidence
• Finding: Saying semantic detection works in any language by default implies broad passive analysis of user content without describing user choice, consent, or any mechanism to restrict monitoring scope. While this is primarily a transparency and consent problem rather than direct code execution risk, it can normalize always-on behavioral inference and expand privacy exposure across all user communications.

Vague Triggers

Medium

• Confidence: 82% confidence
• Finding: The phrase 'when triggered' leaves activation conditions undefined, which can cause the agent to run in broader contexts than users expect. For a proactive skill that inspects user context and may read files, ambiguous triggering increases the risk of unintended monitoring, surprise actions, and privacy-invasive operation.

Missing User Warnings

High

• Confidence: 95% confidence
• Finding: The skill explicitly references reading journals and medical logs, which are highly sensitive categories of personal data, but does not provide a clear privacy warning, consent model, or data-handling limits. In the context of an energy-detection and task-prioritization skill, this makes the issue more dangerous because the functionality encourages ongoing inspection of intimate user state rather than a one-time, clearly bounded read.

VirusTotal

65/65 vendors flagged this skill as clean.