Security audit

Java

Security checks across malware telemetry and agentic risk

Overview

This Java helper skill is coherent and disclosed; it uses a scoped local preferences folder to tailor Java advice, with no evidence of exfiltration or hidden execution.

Installers should know this skill may keep Java-related preferences and recurring context in ~/Clawic/data/java/ and may consult ~/Clawic/profile.yaml for locale/timezone defaults. Review or delete those local files if you do not want cross-session personalization.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (5)

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: The setup instructions direct the agent to read multiple user-scoped files from the home directory, including profile and prior-memory data, without any user-facing notice or consent step. This creates a privacy and transparency issue because the agent may silently ingest sensitive preferences, environment details, or historical context that the user did not expect to be accessed for the current task.

Missing User Warnings

Medium

Confidence: 97% confidence
Finding: The skill instructs the agent to write user preferences and behavioral context to persistent files, but gives no explicit warning that information will be stored across sessions. Silent persistence of user statements can create privacy, retention, and trust problems, especially when storing work habits, safety posture, or other contextual metadata beyond the immediate conversation.

Natural-Language Policy Violations

Low

Confidence: 88% confidence
Finding: Telling the agent not to interview the user and never open with setup questions removes an important consent and transparency checkpoint before reading configuration or memory. In this file, that instruction directly increases the likelihood of silent data access and persistence, making otherwise moderate privacy issues easier to trigger without user awareness.

Ssd 3

Medium

Confidence: 95% confidence
Finding: The instructions explicitly require persistent storage of user-revealed preferences and habits across sessions in local memory files. Cross-session retention of such context can accumulate sensitive behavioral or operational information over time and may be reused in later interactions without fresh consent or visibility.

Ssd 3

Medium

Confidence: 94% confidence
Finding: The agent is instructed to load prior context and recurring pain points from a persistent memory file on startup, which enables silent reuse of historical user information in future sessions. This is risky because users may not expect old context to influence new conversations, and the stored data may contain sensitive technical, organizational, or preference details.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.