Keep Learning Agent

• Outcome: Review
• Latest audit: May 29, 2026 at 9:22 PM UTC
• Version:
  1.0.0

Overview

This is a coherent learning and memory framework, but it asks agents to automatically run unreviewed local PowerShell scripts and persist changes into agent configuration without enough scoping or user control.

Install only if you intentionally want a persistent agent learning system. Before enabling it, inspect or disable the G:\clawbot PowerShell scripts, keep learning files in a scoped directory, avoid storing secrets, and manually review anything before it is promoted into AGENTS.md or other global agent configuration.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• 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
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (3)

Missing User Warnings

Medium

• Confidence: 87% confidence
• Finding: The README instructs users that a PowerShell self-repair script runs automatically at session start and references fixed local paths, but provides no warning about what the script changes, what permissions it needs, or how to review/disable it. In an agent skill context, auto-executed maintenance scripts increase risk because they can modify files or environment state implicitly, and users may adopt the skill without scrutinizing the script's behavior.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The skill explicitly documents an automatic session-start workflow that includes running a local PowerShell script from a fixed Windows path, but provides no safety prompt, trust boundary, or explanation of side effects. In an agent setting, normalizing automatic execution of local scripts can lead to unintended code execution, configuration changes, or persistence behaviors if the referenced script is modified or malicious.

Missing User Warnings

Low

• Confidence: 84% confidence
• Finding: The maintenance section instructs moving files, archiving records, and updating indexes, but does not warn that these actions modify persistent user data. While less severe than code execution, silent data mutation can cause loss of history, incorrect indexing, or unintended overwrites if an agent applies the procedure automatically.

VirusTotal

66/66 vendors flagged this skill as clean.