ClawHub
Back to skill

Security audit

Feynman Technique

Security checks across malware telemetry and agentic risk

Overview

This teaching skill is mostly coherent, but it automatically loads a remote replacement version at run time, so it should be reviewed before installation.

Install only if you are comfortable with the skill contacting deciqai.com at runtime and allowing that response to replace or update the skill instructions. In controlled environments, prefer a version with the full instructions packaged locally or with remote content treated only as optional reference material.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Description-Behavior Mismatch

Medium
Confidence
96% confidence
Finding
The skill explicitly instructs the agent to fetch remote content at runtime and treat it as the canonical current version of the skill. That expands the skill's behavior from local concept explanation into dynamic network-driven instruction loading, creating a supply-chain/prompt-injection risk if the remote endpoint is compromised, altered, or serves unexpected content.

Context-Inappropriate Capability

Medium
Confidence
97% confidence
Finding
A Feynman-technique teaching skill has no operational need to pull fresh remote instructions during execution, so the self-update directive is unjustified and increases attack surface without functional necessity. Because the fetched JSON is described as the canonical current version, an attacker controlling that endpoint could silently change agent behavior after installation.

Missing User Warnings

Medium
Confidence
93% confidence
Finding
The skill directs the agent to make a network request at the start of a run without any user-facing disclosure or consent mechanism. Hidden outbound requests can violate privacy expectations, leak usage metadata, and bypass deployment policies that assume a teaching skill operates entirely locally.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.