Pattern Analyst

• Outcome: Review
• Latest audit: May 28, 2026 at 9:27 AM UTC
• Version:
  1.0.0

Overview

This skill is locally scoped and purpose-aligned, but it can build and update a lasting personal profile from inferred behavior without explicit approval first.

Install only if you want the agent to keep a persistent local pattern log and edit USER.md based on its interpretation of your repeated behavior. Review notes/patterns.md and USER.md regularly, and prefer requiring explicit confirmation before any profile update.

SkillSpector

By NVIDIA

Vulnerability Patterns

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

Findings (6)

Context-Inappropriate Capability

Medium

• Confidence: 94% confidence
• Finding: The skill explicitly instructs the agent to update `USER.md` with inferred profile information, including auto-confirmed patterns derived from repeated behavior. Persisting inferred traits to a user profile without explicit confirmation creates a profiling and privacy risk, and can cause inaccurate or sensitive inferences to become durable across future interactions.

Vague Triggers

Medium

• Confidence: 74% confidence
• Finding: The activation guidance is broad enough that the skill could run during generic heartbeats or after any 'significant' shared content, which is subjective and prone to overuse. Over-broad invocation increases the chance of unnecessary profiling, note-taking, and profile updates in contexts where the user did not clearly request analysis.

Missing User Warnings

Medium

• Confidence: 92% confidence
• Finding: The skill directs the agent to log user-shared content to `notes/patterns.md` without a clear warning, consent flow, or retention boundary. Persistent storage of behavioral analysis can expose private interests, frustrations, and goals across sessions, especially when users may not realize that routine sharing is being recorded.

Missing User Warnings

High

• Confidence: 97% confidence
• Finding: The auto-confirm workflow allows inferred patterns to be written to `USER.md` immediately, without prior user warning or opt-in. This is more dangerous than ordinary note-taking because it promotes probabilistic behavioral inferences into durable profile data that may affect future system behavior and user treatment.

Ssd 3

Medium

• Confidence: 90% confidence
• Finding: The skill is designed to persist and profile cross-conversation interaction content, including inferred intent and goal alignment. Even if meant to help the user, this creates a longitudinal behavioral profile that can accumulate sensitive personal information and amplify privacy harms if inaccurate, excessive, or later reused inappropriately.

Ssd 3

Medium

• Confidence: 96% confidence
• Finding: The auto-confirm workflow converts repeated behavior into persistent profile data without explicit confirmation, effectively allowing the agent to decide what the user's enduring traits or preferences are. This is risky because repeated actions can be misinterpreted, context-dependent, or sensitive, yet the resulting profile changes become durable and influential.

VirusTotal

65/65 vendors flagged this skill as clean.