ClawHub
Back to skill

Security audit

PinkyBrain

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed P2P AI networking tool, but users should treat it as a background network service that can share compute and store conversation data.

Install only if you intentionally want to run a P2P AI node. Review the cloned repository and dependencies before running, keep public mesh participation off unless needed, use a strong P2P secret, restrict shared content to shared_models/, avoid sensitive data in persisted conversations, and pin PINKYBRAIN_URL instead of relying on network auto-discovery in sensitive environments.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • 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
90% confidence
Finding
The README prominently encourages joining a public mesh, sharing compute, and exposing distributed memory/models, but it does not provide an explicit warning about privacy, data exposure, billing/resource consumption, or the risks of participating in an untrusted peer network. In a skill context, this omission is security-relevant because users may enable networked sharing features under the assumption that the advertised protections fully eliminate operational risk, when they do not.

Missing User Warnings

Medium
Confidence
94% confidence
Finding
The installation and quick-start sections instruct users to start a daemon in the background and then join the public mesh without an adjacent caution about persistent background execution, automatic updates, peer discovery, or network participation. This is dangerous because it can cause users to unknowingly expose services, consume local resources, and connect to external peers before they understand the trust and security implications.

Missing User Warnings

Medium
Confidence
85% confidence
Finding
The skill documents persistent conversation storage, search, export, and privacy-level changes, but it does not prominently warn users that prompts and responses may be retained on disk for 90 days and may later be exported or shared. In a distributed AI/memory-sharing context, this creates a real privacy risk because users may submit sensitive data assuming ephemeral processing when the system actually persists it.

Missing User Warnings

Medium
Confidence
92% confidence
Finding
The auto-discovery flow states that the client reads local discovery sources and uses the first responding endpoint, including localhost, Tailscale peers, and mDNS, without a strong warning or trust-verification requirement at selection time. This can expose users to connecting to an unintended or malicious local-network service that impersonates a PinkyBrain node, especially in environments where mDNS or peer-discovery is not tightly controlled.

VirusTotal

1/64 vendors flagged this skill as malicious, and 63/64 flagged it as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.