Llm Safe Write

• Outcome: Pass
• Latest audit: Jun 5, 2026 at 3:53 AM UTC
• Version:
  0.1.0

Overview

This is a self-contained file-writing workflow skill with no evidence of hidden credential use, network behavior, persistence beyond installation, or destructive actions.

Install this only if you want your agent to use a more cautious incremental editing workflow for large or CJK-containing files. Review generated edits as usual, especially append operations that may temporarily insert marker lines if interrupted.

SkillSpector

By NVIDIA

Vulnerability Patterns

• 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
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (1)

Intent-Code Divergence

Medium

• Confidence: 93% confidence
• Finding: The skill contains conflicting operational guidance: earlier sections require proactively avoiding Write for files over 50 lines, CJK content, or after a Write failure, but the Strategy section says to 'try the simpler approach first.' In an agent setting, contradictory instructions can cause unsafe routing, leading the model to retry or attempt Write in cases the skill otherwise classifies as high-risk, increasing the chance of truncated or corrupted file writes.

VirusTotal

64/64 vendors flagged this skill as clean.