ClawHub
Back to skill

Security audit

Pattern Analyst

Security checks across malware telemetry and agentic risk

Overview

This skill is locally scoped and not malicious, but it can build and modify a persistent personal profile from inferred behavior without clear approval before each update.

Install only if you want the agent to keep a persistent local pattern log and edit USER.md based on its interpretation of repeated behavior. Review notes/patterns.md and USER.md regularly, and require explicit confirmation before any profile update if you do not want inferred traits or preferences stored automatically.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (5)

Description-Behavior Mismatch

Medium
Confidence
92% confidence
Finding
The skill is presented as an analysis tool, but it also mutates persistent profile data in USER.md based on inferred patterns. That creates an integrity and privacy risk because inferred or incorrect conclusions can be stored as facts and later influence future agent behavior without explicit per-update user approval.

Vague Triggers

Medium
Confidence
84% confidence
Finding
The activation criteria are broad enough to trigger on ordinary conversation, heartbeats, or any significant shared content, which can cause the skill to run frequently without a strong user intent signal. In this skill's context, overbroad triggering is more dangerous because execution can lead to persistent logging and profile updates from casual interactions.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill directs writes to persistent files, including USER.md, without requiring a user-facing notice or consent at the time of update. This creates covert persistence of behavioral inferences, which can surprise the user, entrench inaccuracies, and expand the amount of sensitive profile data stored over time.

Autonomous Decision Making

Medium
Category
Excessive Agency
Content
Observe → Surface → Enzo confirms → Update USER.md

### 2. Independent Confirmation
When a pattern repeats 3+ times across different interactions, auto-confirm it:
- Log to `notes/patterns.md` with `[AUTO-CONFIRMED]` tag
- Update `USER.md` immediately
- Mention it next conversation: "I've added X to your profile based on repeated behavior"
Confidence
96% confidence
Finding
auto-confirm

Autonomous Decision Making

Medium
Category
Excessive Agency
Content
### 2. Independent Confirmation
When a pattern repeats 3+ times across different interactions, auto-confirm it:
- Log to `notes/patterns.md` with `[AUTO-CONFIRMED]` tag
- Update `USER.md` immediately
- Mention it next conversation: "I've added X to your profile based on repeated behavior"
Confidence
95% confidence
Finding
AUTO-CONFIRM

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.