ClawHub
Back to skill

Security audit

Karpathy LLM Wiki

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed personal wiki workflow that writes only to a scoped workspace knowledge-base directory and does not show deceptive or destructive behavior.

Install this if you want an agent-managed workspace wiki. Expect it to create and update wiki-kb/ files during ingest, archive, and lint operations, and to fetch source material when you ask it to ingest external content. Review lint auto-fixes if preserving exact wiki structure matters to you.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Vague Triggers

Medium
Confidence
91% confidence
Finding
The trigger list is broad and includes common phrases like 'what do I know about' and generic references to querying or adding to a knowledge base. This can cause the skill to activate in situations where the user did not intend wiki reads or writes, increasing the chance of unintended file operations or external fetches.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill performs filesystem creation and modification, including creating directories and writing index and log files, but the description does not prominently warn users that invoking the skill may change workspace contents. In a conversational environment, missing disclosure raises the risk of surprising or unauthorized writes when the skill is auto-selected.

Missing User Warnings

Low
Confidence
92% confidence
Finding
The ingest workflow instructs the agent to fetch external content using available web or file tools, but the skill description does not clearly disclose that network or external-source access may occur. This can lead to privacy, provenance, or SSRF-like concerns depending on the host environment and available connectors.

VirusTotal

62/62 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.