Self-Evolving

• Outcome: Pass
• Latest audit: May 28, 2026, 10:20 AM
• Version:
  1.0.0

Overview

This markdown-only skill is coherent with its stated purpose of improving repeated workflows through local notes, with no scripts, network calls, credential access, or hidden execution.

Install this if you want the agent to keep concise local notes about repeated workflows and improvement experiments. Review ~/self-evolving/ periodically, avoid saving secrets or sensitive personal/business details there, and tell the agent to pause or opt out if you do not want ongoing memory updates.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• 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

Findings (4)

Missing User Warnings

Medium

• Confidence: 89% confidence
• Finding: The template explicitly instructs the agent to create and continually update a persistent file under the user's home directory, which is a local data modification action. Even though the content appears operational rather than overtly malicious, it lacks any user-consent, disclosure, or safety constraints around when filesystem writes are permitted, which can lead to unauthorized persistence, privacy issues, or unintended state changes.

Vague Triggers

Medium

• Confidence: 85% confidence
• Finding: The activation guidance is intentionally broad and asks the agent to infer when to activate from repeated workflows, friction, or optimization topics within the first few exchanges. Without hard boundaries, the skill may engage in contexts the user did not clearly authorize, increasing the chance of unexpected behavior changes or unsolicited collection of preferences tied to ongoing work.

Ssd 3

Medium

• Confidence: 93% confidence
• Finding: The skill explicitly directs the agent to persist user-derived information into memory files and archives for future reuse, but it does not limit what kinds of data may be stored or exclude sensitive content. In practice, this can lead to retention of confidential preferences, workflow details, mistakes, or other personal/business information beyond the immediate conversation need.

Ssd 3

Medium

• Confidence: 78% confidence
• Finding: Telling the agent to keep learning through normal use and reflect understanding after each answer creates pressure to continuously extract and accumulate user-derived signals over time. Even without explicit logging language, this encourages implicit profiling and retention creep, especially when paired with the earlier instructions to maintain local notes and experiment history.

VirusTotal

VirusTotal findings are pending for this skill version.