Security audit

Wx Peitu

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed WeChat illustration generator that creates local PNGs and syncs them to Feishu, with automatic side effects users should understand before use.

Install only if you are comfortable with generated article-derived images being saved to your Desktop and uploaded to Feishu/Lark Drive using your local lark-cli session. Use the confirmation-based flow for sensitive articles, remove or disable the upload/open-folder steps if you want local-only output, and periodically clear the photo-deduplication file if you do not want cross-run state.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
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

Findings (7)

Scope Creep

Medium

Confidence: 96% confidence
Finding: The workflow introduces a persistent global file, references/used-photos.json, to track previously used photo IDs across runs and articles. That exceeds the skill's declared behavior of reading a local MD file and writing output artifacts, and it creates undeclared cross-session state that can surprise users, leak usage patterns, and complicate data lifecycle expectations.

Intent-Code Divergence

Medium

Confidence: 89% confidence
Finding: The documented requirement says images should be pre-downloaded locally for reliability, but the template example still uses an external Unsplash URL. In a workflow that later launches Chrome with web security disabled, contradictory guidance increases the chance that generated HTML will fetch remote content at render time, causing privacy leakage, nondeterministic output, and unexpected network access.

Vague Triggers

Medium

Confidence: 82% confidence
Finding: The trigger phrases are broad enough that normal conversation about public-account articles or illustration requests could invoke the skill unintentionally. In this skill's context, accidental invocation matters because the documented workflow includes local file reads, external API calls, desktop writes, possible explorer execution, and automatic Feishu cloud uploads, so a false trigger can cause real side effects and data egress.

Vague Triggers

Medium

Confidence: 91% confidence
Finding: The keyword shortcut table maps broad, everyday phrases like “排版”, “清单”, “步骤”, and “vs” directly to layout/mode/theme behavior. In an agent skill, overly broad automatic triggers can cause the skill to activate or steer execution when the user did not intend to invoke this specific behavior, increasing the risk of unintended file reads, image generation, subprocess use, or cloud sync within the skill’s declared capabilities.

Vague Triggers

Medium

Confidence: 84% confidence
Finding: The trigger phrases include generic terms like '多图', '配图', and '全套', which are broad enough to match ordinary conversation unrelated to this skill's higher-risk capabilities. Because the skill reads local files, invokes subprocesses, and syncs to cloud storage, accidental activation can lead to unintended file access, network calls, and artifact upload.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The workflow defines cross-article persistent photo blacklist writes and later cloud sync behavior without clear user-facing warning or consent at the point of action. In this skill context, undisclosed persistence and external upload are especially risky because they create durable state and move generated outputs off-device, which can violate user expectations and privacy constraints.

Autonomous Decision Making

Medium

Category: Excessive Agency
Content: ## ⚠️ User Warnings This skill automatically performs the following operations when running: 1. **Auto-uploads to Feishu Cloud Drive**: Generated illustrations are automatically uploaded to a specified Feishu cloud folder 2. **Auto-saves to Desktop**: PNG illustrations are automatically saved to `C:\Users\<user>\Desktop\<project-name>公众号配图\` directory 3. **Auto-calls external APIs**: Calls Pexels/Pixabay APIs to search for illustration photos (requires API Key)
Confidence: 93% confidence
Finding: automatically perform

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.