Translation Assistant

• Outcome: Review
• Latest audit: Jun 2, 2026, 10:22 AM
• Version:
  1.0.0

Overview

This skill is mostly non-executable documentation, but it mixes a translation assistant with unrelated Home Assistant plugin-installation instructions and very broad triggers, so users should review it before installing.

Install only if you are comfortable with a messy, auto-merged skill file. It does not show malicious code or hidden data access, but the unrelated Home Assistant sections and broad "home" trigger should be cleaned up to avoid accidental activation or confusing the agent about what the skill is meant to do.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• 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

Findings (6)

Description-Behavior Mismatch

Medium

• Confidence: 95% confidence
• Finding: The skill declares itself as a translation assistant, but large merged sections document unrelated Home Assistant plugin-installation content. This creates capability confusion and can cause the agent to respond or route requests outside the declared purpose, increasing the chance of unintended behavior and misuse.

Intent-Code Divergence

Medium

• Confidence: 97% confidence
• Finding: This block explicitly re-labels the translation skill as a Home Assistant plugin-installation skill, directly contradicting the manifest and earlier documentation. Such contradictory instructions can mislead dispatch, operators, or downstream systems into invoking the skill for unintended domains, undermining trust boundaries and safe routing.

Intent-Code Divergence

Medium

• Confidence: 97% confidence
• Finding: A second contradictory merged section repeats the Home Assistant plugin-installation framing, reinforcing behavior drift away from the declared translation purpose. Repeated divergence suggests broken content-merging hygiene and increases the likelihood that the skill will activate or be trusted in the wrong context.

Vague Triggers

High

• Confidence: 91% confidence
• Finding: The trigger phrases include broad everyday requests like asking to translate a passage or how to translate, which can cause accidental invocation in normal conversation rather than explicit skill activation. In an agent system, overbroad triggers can route user input into the wrong capability, causing unexpected actions, privacy issues, or policy bypass through misclassification.

Vague Triggers

High

• Confidence: 99% confidence
• Finding: Using the generic term "home" as a trigger is dangerously broad and likely to match many unrelated user utterances. Because this term appears inside a mislabeled merged section, it further increases the chance that unrelated conversations will invoke the wrong skill or capability path.

Vague Triggers

High

• Confidence: 99% confidence
• Finding: The repeated use of "home" as a trigger in another merged section compounds the accidental invocation risk and shows the problem is systemic, not isolated. Repetition across merged blocks makes the skill especially unsafe because multiple documentation paths instruct activation from an extremely common word.

VirusTotal

VirusTotal findings are pending for this skill version.