ClawHub

Nutrition Claw

Security checks across malware telemetry and agentic risk

Overview

This appears to be a legitimate local nutrition tracker, but it needs review because malformed date inputs can make it read or write files outside its intended nutrition folder.

Review before installing. The skill appears designed to keep data local and I found no exfiltration or destructive intent, but it should validate date inputs before reading or writing logs. If you use it, pass only normal YYYY-MM-DD dates, avoid running it with elevated privileges, and remember that your nutrition history is stored locally in plaintext under ~/.nutrition-claw/.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (4)

Missing User Warnings

Medium
Confidence
97% confidence
Finding
The skill instructs the agent to create and maintain `~/.nutrition-claw/feedback-diary.txt`, a new persistent store of user food history and inferred educational metadata, without clear user consent, retention notice, or privacy warning. Because food logs can reveal sensitive health, lifestyle, and potentially medical information, this creates an avoidable privacy and data-governance risk even though the storage is local.

Natural-Language Policy Violations

High
Confidence
98% confidence
Finding
The skill requires the agent to give confident, unhedged nutrition and health commentary framed as factual science, including prescriptive coaching tied to user goals. That is dangerous because LLM-generated health guidance may be inaccurate, overgeneralized, or inappropriate for users with medical conditions, disordered eating history, or special populations, yet the instructions discourage caution and uncertainty.

Natural-Language Policy Violations

High
Confidence
99% confidence
Finding
The topic list directs the agent to discuss disease risk, hormones, blood pressure, insulin sensitivity, liver health, mental health, immune function, and similar medical/scientific subjects without any validation or safeguards. This meaningfully increases the chance of harmful misinformation or quasi-medical advice being delivered as personalized guidance from ordinary meal logging interactions.

Ssd 3

Medium
Confidence
97% confidence
Finding
The skill mandates persistent logging of user food inputs plus derived topic metadata in a diary file, which expands the user's stored behavioral profile beyond the main nutrition logs. Even if local-only, this creates additional sensitive data surface, longer retention, and potential secondary use without clear transparency, access controls, or deletion UX.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal