Security audit

日志压缩器免费版

Security checks across malware telemetry and agentic risk

Overview

This skill is a log-compression helper, but its documentation is inconsistent about whether logs stay local and it directs persistent memory changes without clear confirmation boundaries.

Review this skill carefully before installing. Use it only on logs you are willing for your agent or model provider to process, verify the actual compression script exists before running any command, and manually inspect the generated summary before appending anything to MEMORY.md.

SkillSpector

Vulnerability Patterns

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (5)

Intent-Code Divergence

Medium

Confidence: 92% confidence
Finding: The documentation makes conflicting security-relevant claims: it says the skill is a pure local Node.js script with no external dependencies or API key, while elsewhere it declares a required LLM API. This can mislead users and reviewers about data flow, causing them to expose logs or sensitive memory content under the false assumption that processing never leaves the local environment.

Intent-Code Divergence

Medium

Confidence: 95% confidence
Finding: The skill simultaneously claims no API key is needed because processing is local and also lists an external LLM API as required. This inconsistency is dangerous because operators may permit execution in sensitive environments without realizing an external service dependency could receive the log data.

Vague Triggers

Medium

Confidence: 86% confidence
Finding: The activation language is broad enough to match generic development or memory-compression scenarios rather than a narrowly scoped, user-confirmed action. In an agent environment with read/exec/write capabilities, vague triggering increases the chance the skill runs unexpectedly on arbitrary files or context, leading to unintended command execution or data modification.

Vague Triggers

Medium

Confidence: 90% confidence
Finding: The example trigger phrase is ordinary conversational language ('今天的日志太长了，帮我压缩') and can easily appear in normal chat. In a skill with execution and file-write behavior, such a broad trigger can cause accidental activation and processing of logs or persistent files without sufficiently explicit consent.

Missing User Warnings

Medium

Confidence: 97% confidence
Finding: The instructions direct appending generated output into MEMORY.md, which is persistent user data, without an explicit warning, review gate, or consent step. This is risky because compressed or malformed output could silently pollute long-term memory/state, propagate errors, or preserve sensitive log content in a durable store.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.