Security audit

Translate

Security checks across malware telemetry and agentic risk

Overview

This translation skill is coherent and disclosed, but it does keep persistent local notes and users should understand that before installing.

Install only if you want this skill to maintain local translation memory under ~/Clawic/data and related shared contacts, projects, and finance notes. For confidential, personal, legal, or medical text, use its non-hosted workflow guidance and do not approve hosted MT unless the provider terms are acceptable.

SkillSpector

By NVIDIA

Vulnerability Patterns

Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (6)

Description-Behavior Mismatch

Medium

Confidence: 94% confidence
Finding: The skill asserts that 'nothing leaves the machine' while elsewhere explicitly allowing machine-translation workflows that may send user content to third-party services. That mismatch can mislead users and downstream agents about confidentiality boundaries, causing sensitive text to be processed externally without informed consent.

Intent-Code Divergence

Medium

Confidence: 97% confidence
Finding: The document contains contradictory privacy guarantees: one section says nothing leaves the machine, while another says MT may send content to third parties. In a translation skill, this contradiction is security-relevant because users may provide NDA, legal, medical, or personal material and rely on the stronger claim.

Context-Inappropriate Capability

Medium

Confidence: 92% confidence
Finding: The skill instructs the agent to read a local file from a user-home path (`~/Clawic/data/translate/memory.md`) before performing work. For a translation skill, accessing persistent local memory is not necessary to translate text and creates an unnecessary data-access channel that could expose unrelated local information or let skill content influence behavior via hidden state. The skill context makes this more dangerous because it normalizes consulting local files as part of ordinary translation/debugging workflow.

Context-Inappropriate Capability

Medium

Confidence: 96% confidence
Finding: The file instructs writing notes and artifacts to persistent local paths such as `~/Clawic/data/translate/memory.md` and `artifacts/<surface>-rendering.md`. A translation/localization skill does not inherently need filesystem write capability, so these directions can cause unauthorized persistence, contamination of local state, and leakage of user or incident data into files outside the immediate task. In this context, the danger is elevated because the writes are framed as routine documentation steps rather than an explicitly approved action.

Missing User Warnings

Low

Confidence: 88% confidence
Finding: The markdown tells the agent to write incident and environment details to files in the user home area without any warning or confirmation step. Even if the content is operationally benign, silently modifying local files violates least surprise and can create privacy, audit, and integrity issues by persisting data the user did not authorize. The translation skill context makes this less justified, since users would not expect filesystem side effects from translation guidance.

Missing User Warnings

Low

Confidence: 91% confidence
Finding: The instruction explicitly directs the user to access client-specific translation memory and glossary data from a local filesystem path, but provides no accompanying confidentiality, least-privilege, or handling guidance. In a translation workflow, those assets often contain proprietary text, customer terminology, or regulated content, so normalizing direct access without privacy controls increases the risk of unnecessary exposure or misuse.

VirusTotal

62/62 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.