ClawHub
Back to skill

Security audit

Cohere Translator

Security checks across malware telemetry and agentic risk

Overview

This is a coherent Cohere translation skill, but anything translated, including files, is sent to Cohere's external API.

Install only if you are comfortable sending translated text or selected files to Cohere. Avoid using it for secrets, credentials, regulated data, or confidential documents unless your organization approves that data flow, and prefer COHERE_API_KEY in the environment rather than passing keys in shared command history.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (6)

Intent-Code Divergence

Medium
Confidence
96% confidence
Finding
The documentation states that file mode 'bypasses agent context entirely,' which can mislead users into believing the file stays local, but the implementation uploads the file contents to Cohere for translation. In a security-sensitive agent skill, this mismatch materially increases the risk of accidental disclosure of sensitive data to an external service.

Intent-Code Divergence

Medium
Confidence
97% confidence
Finding
The CLI help/comment claims file mode uses 'ZERO agent tokens' and 'bypasses agent context,' but this framing obscures that the file is still exfiltrated to a third-party API. In an agent setting, such wording can encourage users to route sensitive files through the tool under a mistaken privacy assumption.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The README explicitly promotes a mode where file contents are sent to Cohere for translation, but it does not clearly warn users that potentially sensitive local document contents will be transmitted to a third-party service. In an agent setting, this increases the risk of accidental disclosure of confidential data because users may interpret 'the agent never touches the content' as implying privacy or safety when the content is still exfiltrated off-host to an external API.

Missing User Warnings

Medium
Confidence
96% confidence
Finding
The skill strongly encourages file-based translation and emphasizes that the agent should bypass reading file contents, but it does not clearly warn users that the file contents are still transmitted to Cohere's external API. This can mislead users into believing file mode is privacy-preserving when it only reduces agent token usage, creating a meaningful risk of unintended third-party data disclosure.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The script transmits user-supplied text to Cohere's external API, but the user-facing interface emphasizes token savings rather than clearly warning that data leaves the local environment. In a skill context, this can cause inadvertent sharing of secrets, proprietary text, or regulated data.

External Transmission

Medium
Category
Data Exfiltration
Content
The script sends a chat completion request to Cohere's API:

```
POST https://api.cohere.ai/v2/chat
Authorization: Bearer $COHERE_API_KEY
{
  "model": "command-a-translate-08-2025",
Confidence
93% confidence
Finding
https://api.cohere.ai/

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.