Security audit

WhatsApp表情搜索

Security checks for vulnerabilities and agentic risk

Overview

The skill appears to be a WhatsApp GIF sender, but its routing text also claims unrelated SEO use cases that could cause it to run in the wrong context.

Review carefully before installing. Use this only for explicit WhatsApp GIF search, conversion, and sending tasks, confirm the recipient and file before any send action, and do not rely on it for SEO or generic automation claims. The artifact does not show malware, but the mismatched routing language and external message-sending capability warrant review.

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 (5)

Description-Behavior Mismatch

High

Confidence: 98% confidence
Finding: The manifest description conflates a WhatsApp GIF-sending utility with unrelated SEO optimization use cases, which can cause the agent to invoke the skill in inappropriate contexts. This expands the skill’s activation surface and may lead to unintended command execution or media-sending behavior when a user asked for SEO help rather than messaging automation.

Intent-Code Divergence

Medium

Confidence: 93% confidence
Finding: The skill claims coverage for capabilities and keywords that are not actually implemented by the documented behavior. Misrepresenting coverage can mislead an agent into trusting the skill for tasks outside its real scope, increasing the chance of unsafe or unintended execution paths.

Vague Triggers

High

Confidence: 98% confidence
Finding: The trigger conditions are unrelated to the skill’s real WhatsApp GIF purpose and are written broadly enough to match SEO requests. In an agent environment, this can cause the wrong skill to activate and perform shell-based download/convert/send actions in response to unrelated prompts.

Vague Triggers

Medium

Confidence: 95% confidence
Finding: The manifest uses broad invocation language without clear boundaries, making it easier for the orchestration layer to select this skill for tasks it does not actually implement. Overbroad routing increases the risk of accidental command execution and unintended external actions such as downloading files or sending media.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The skill automates sending downloaded media to WhatsApp contacts but does not prominently warn that content will be transmitted to an external recipient and service. Without explicit consent and transmission disclosure, users may unintentionally send media to the wrong contact or expose content outside the local environment.

Static analysis

No suspicious patterns detected.