Security audit

表情符号工具箱

Security checks across malware telemetry and agentic risk

Overview

This skill needs review because it teaches hiding messages and digital-token strings inside emoji for public chats, with limited guardrails and some mismatched scope and offline claims.

Install only if you intentionally need authorized Unicode steganography, watermark testing, or defensive analysis. Do not use it to hide secrets, credentials, financial tokens, or messages from people who have a right to see or moderate the communication; prefer standard encrypted and authenticated channels for sensitive transfers.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
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
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (8)

Intent-Code Divergence

Medium

Confidence: 89% confidence
Finding: The skill claims operations are local/offline, yet elsewhere instructs users to verify token validity by contacting a mint server. This creates a misleading trust boundary: operators may assume no network exposure while the documented workflow can still drive outbound communication, increasing privacy, telemetry, and data-leak risk.

Intent-Code Divergence

Medium

Confidence: 96% confidence
Finding: The documentation says the skill does not require networking, but the error-handling section instructs the agent to run ping and check firewall/proxy connectivity. In an agent context with exec capability, this can trigger unintended network probing or environment discovery contrary to the stated operating model.

Vague Triggers

High

Confidence: 91% confidence
Finding: The scope text includes unrelated broad trigger phrases such as data analysis, report generation, and visualization, which do not match the emoji toolkit domain. Overbroad activation criteria can cause the agent to invoke this skill in inappropriate contexts, exposing exec/write capabilities and hidden-message logic where they are not relevant.

Missing User Warnings

Medium

Confidence: 84% confidence
Finding: The skill introduces covert hidden-message transmission as a core feature before clearly warning about misuse, detectability, compliance, and privacy risks. Presenting stealth communication as a normal workflow lowers user friction for deceptive use and fails to establish safety constraints at the point of first exposure.

Ssd 2

Medium

Confidence: 95% confidence
Finding: The skill explicitly promotes hiding messages inside emoji using invisible Unicode variation selectors so information is not immediately noticed in public channels. In context, this is covert obfuscation guidance that can facilitate stealthy exfiltration, policy evasion, or concealed transfer of sensitive data.

Ssd 2

Medium

Confidence: 94% confidence
Finding: The guidance teaches users how to make hidden content less noticeable and more resilient in transit, including carrier selection and adding ordinary text to reduce truncation. These are operational stealth-improvement techniques that materially increase the effectiveness of covert communication.

Ssd 4

Medium

Confidence: 93% confidence
Finding: The scenario normalizes using hidden emoji messages in public group chats specifically to avoid attracting attention. This is a concrete deception use case that encourages concealed communication in shared spaces, increasing misuse potential beyond abstract technical explanation.

Ssd 2

Medium

Confidence: 92% confidence
Finding: This scenario promotes concealing token strings during chat transmission to reduce their visibility to others, operationalizing hidden transfer of valuable digital assets. Because the skill also mentions local metadata parsing and transmission reliability, the context makes the concealment workflow more actionable and risky.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.