Cat Therapy

• Outcome: Review
• Latest audit: May 29, 2026, 9:24 PM
• Version:
  1.0.1

Overview

This cat-relaxation skill is mostly coherent, but it ships with leftover user-specific preference data and stores custom photos or sounds in a shared local file with weak privacy boundaries.

Review or delete user_cats.json before installing. Only use custom photo or sound features if you are comfortable with that media being stored in the skill directory and reused in chat platforms; check how your OpenClaw environment scopes skill files between users or workspaces.

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 (6)

Vague Triggers

High

• Confidence: 94% confidence
• Finding: The Chinese trigger phrases are very broad everyday expressions such as '休息一下' and '累了', which can easily appear in normal conversation without the user intending to invoke this skill. In a multi-skill or always-listening assistant environment, this can cause unsolicited activation, unexpected outbound messages, and reduced user control over when the skill runs.

Vague Triggers

High

• Confidence: 95% confidence
• Finding: The English triggers include common phrases like 'tired', 'stress', and 'relax', which are likely to occur in ordinary chat and are not unique commands. This makes accidental triggering more likely across supported messaging platforms, especially because the skill is designed to auto-send media and audio as a response.

Missing User Warnings

Medium

• Confidence: 92% confidence
• Finding: The README says user-uploaded cat photos and audio can be saved and reused, but gives no notice about storage, retention period, deletion behavior, or who can access the content. Because these are user-provided media files and voice samples that may be personally identifying, the lack of transparency and consent controls creates privacy and compliance risk.

Vague Triggers

Medium

• Confidence: 89% confidence
• Finding: The trigger design is overly broad for a cross-platform auto-send skill, so ordinary phrases like being tired or wanting to relax can unintentionally activate external media sending and preference-related flows. In chat environments this can cause unexpected outbound actions, accidental data processing, and user confusion at scale.

Vague Triggers

Medium

• Confidence: 95% confidence
• Finding: Single-word triggers such as 'stress', 'relax', and '喵' are highly ambiguous and likely to appear in normal conversation, making accidental activation even more probable. Because the skill can send media, call TTS, and use stored preferences, this ambiguity increases the chance of unintended processing or transmission.

Missing User Warnings

Medium

• Confidence: 91% confidence
• Finding: The skill supports user-uploaded images and multi-platform delivery, but it does not clearly warn users that their provided media may be transmitted to external platforms or fetched from external services. This omission undermines informed consent and can expose personal photos or identifiers beyond what a user expects from a casual wellness feature.

VirusTotal

66/66 vendors flagged this skill as clean.