Active Self-Improvement

• Outcome: Review
• Latest audit: May 29, 2026 at 9:30 PM UTC
• Version:
  1.3.0

Overview

This skill is transparent about its goal, but it lets an agent automatically change its own skills, memory, and behavior rules without enough user control.

Install only if you intentionally want an agent to revise its own instructions and memory. Use dry-run mode, review diffs before writes, and avoid weekly or automatic runs unless you define exactly which files it may read and change.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (2)

Vague Triggers

High

• Confidence: 98% confidence
• Finding: The skill is configured to trigger on broad, ambient conditions such as end of sessions, batch completion, project milestones, explicit improvement prompts, and a weekly cron schedule. That makes privileged self-modification behavior run without a narrowly scoped user request, increasing the chance of unauthorized or context-inappropriate changes to skills, behavior, or memory.

Missing User Warnings

High

• Confidence: 99% confidence
• Finding: The skill explicitly states it will update skills, protocols, behavior, and memory automatically, and later says low/medium-risk changes may be applied immediately or with delayed notification. This creates a self-modifying loop with no upfront user warning or consent, which is dangerous because poisoned logs, incorrect inferences, or adversarial content in learnings/memory can silently alter future agent behavior and persistence layers.

VirusTotal

59/59 vendors flagged this skill as clean.