Omem

• Outcome: Review
• Latest audit: Jun 5, 2026 at 3:52 AM UTC
• Version:
  1.0.0

Overview

This skill is not malicious, but it gives agents broad access to sensitive local work memory with an overbroad default-to-search trigger.

Review before installing if your OMem index includes confidential company, client, HR, legal, financial, or personal material. The skill is read-oriented and explicitly avoids setup or mutation commands, but it encourages agents to search private work memory on broad or implicit work-related prompts. Install only if you are comfortable with that default, and prefer using it with explicit prompts such as 'search my OMem for...' when sensitive data may be involved.

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 (4)

Vague Triggers

High

• Confidence: 94% confidence
• Finding: The skill directs the agent to use OMem whenever a question touches work context 'even implicitly' and to query first when in doubt. Because OMem spans highly sensitive local sources like email, calendars, documents, and notes, this broad trigger can cause unnecessary access to private data for borderline or ambiguous prompts, violating data minimization and increasing the chance of oversharing.

Vague Triggers

Medium

• Confidence: 88% confidence
• Finding: The secondary guidance repeats broad activation rules such as triggering on many loosely defined work-related categories and 'when in doubt, call it.' This reinforces a default-to-search behavior that can expand retrieval beyond necessity, causing the agent to inspect sensitive workplace memory stores without sufficient user confirmation.

Missing User Warnings

High

• Confidence: 97% confidence
• Finding: The description emphasizes broad access to locally indexed emails, calendars, documents, and collaboration notes but does not prominently warn that these sources may contain confidential or sensitive information. Without a clear privacy notice and consent-oriented usage guidance, the skill encourages agents to search a large corpus of private work artifacts without adequately signaling the sensitivity of that action.

Missing User Warnings

Medium

• Confidence: 92% confidence
• Finding: The documentation explicitly instructs agents to retrieve absolute source file paths and parsed content derived from local files, emails, calendars, and collaboration data. In this skill’s context, that materially increases the chance of exposing sensitive local-path metadata and full parsed document contents to the model or downstream responses without an explicit user-consent, minimization, or warning boundary.

VirusTotal

64/64 vendors flagged this skill as clean.