语义压缩

• Outcome: Pass
• Latest audit: May 29, 2026 at 8:37 PM UTC
• Version:
  1.0.1

Overview

This is a small local prompt-building tool for text compression; its main risk is overpromising lossless compression, not hidden access or malicious behavior.

Install only if you are comfortable with a best-effort prompt helper rather than a guaranteed lossless compressor. Review any compressed output before relying on it, and redact secrets or personal data before saving conversation history to long-term memory.

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 (4)

Intent-Code Divergence

Medium

• Confidence: 90% confidence
• Finding: The prompt makes an absolute guarantee of preserving all key information and forbids summarization, but later instructs the model to compress older dialogue by keeping only conclusions, decisions, and key information. That contradiction can cause silent loss of context, assumptions, constraints, or safety-relevant details from earlier turns, which is dangerous when the compressed output is reused as model context for later decisions.

Description-Behavior Mismatch

Medium

• Confidence: 95% confidence
• Finding: The implementation does not perform semantic compression; it only prepends a fixed instruction prompt to the original text and returns or writes that prompt. This is dangerous because downstream systems may trust the output as already-compressed data, causing prompt bloat, unexpected token consumption, leakage of full original content, and incorrect security or cost assumptions.

Intent-Code Divergence

Medium

• Confidence: 91% confidence
• Finding: The documentation claims accuracy-preserving semantic compression, but the code never compresses content. Misrepresenting behavior at the module level is security-relevant because integrators may rely on the stated guarantees when handling sensitive or large context, leading to data retention, over-disclosure, or failure of controls that assume reduction has occurred.

Missing User Warnings

Medium

• Confidence: 88% confidence
• Finding: The skill explicitly recommends compressing and saving conversation history to long-term memory, but it provides no warning about retaining potentially sensitive user data, privacy implications, or consent requirements. In a context-management skill, this omission can lead operators to persist personal, confidential, or regulated data longer than intended, increasing privacy and compliance risk.

VirusTotal

38/38 vendors flagged this skill as clean.