Back to skill

Security audit

HTML to WeChat

Security checks for vulnerabilities and agentic risk

Overview

This skill is for publishing HTML to WeChat, but it understates conversion and URL-fetching behavior while relying on unpinned installs and WeChat credentials.

Review this before installing. Use it only if you are comfortable giving the workflow access to WeChat Official Account credentials and uploading article content to WeChat. Install dependencies and the companion skill deliberately, prefer pinned versions, keep secrets out of chats and source files, and review the generated HTML, cover image, title, and digest before allowing the final draft upload.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T08 · Insecure Dependencies

Warning
Location
scripts/compress_image.py:27
Finding

Unpinned Dependencies and Automatic Runtime Package Installation

Content
View full analysis
All pip dependencies auto-install on first use if missing. ``` ### Technical Analysis The bundled image-compression script automatically invokes `pip` when Pillow is unavailable. It does not specify an exact version, verify a package hash, use a lockfile, or require approval immediately before installation. Consequently, the package version resolved at execution time may differ from the version that was reviewed. The setup documentation similarly installs multiple mutable package names and an unversioned companion Skill. The companion Skill is particularly security-sensitive because it supplies the conversion and publishing scripts and is expected to process article content and access WeChat credentials. Its source was not included in this project, so its integrity and behavior could not be verified during this audit. Pack ...[truncated 1902 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (21)

Tp2

High
Category
MCP Tool Poisoning
Confidence
85% confidence
Finding

Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.

Content

No source excerpt is available for this finding.

Tp2

High
Category
MCP Tool Poisoning
Confidence
85% confidence
Finding

Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.

Content

No source excerpt is available for this finding.

Tp2

High
Category
MCP Tool Poisoning
Confidence
85% confidence
Finding

Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.

Content

No source excerpt is available for this finding.

Tp2

High
Category
MCP Tool Poisoning
Confidence
85% confidence
Finding

Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.

Content

No source excerpt is available for this finding.

Tp2

High
Category
MCP Tool Poisoning
Confidence
85% confidence
Finding

Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.

Content

No source excerpt is available for this finding.

Tp2

High
Category
MCP Tool Poisoning
Confidence
85% confidence
Finding

Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.

Content

No source excerpt is available for this finding.

Tp2

High
Category
MCP Tool Poisoning
Confidence
85% confidence
Finding

Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.

Content

No source excerpt is available for this finding.

Tp2

High
Category
MCP Tool Poisoning
Confidence
85% confidence
Finding

Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.

Content

No source excerpt is available for this finding.

Tp2

High
Category
MCP Tool Poisoning
Confidence
85% confidence
Finding

Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The manifest promises 'no file conversion' and 'no web scraping,' but the workflow fetches arbitrary URLs, rewrites HTML, compresses images, and publishes to a third-party account. This mismatch is security-relevant because users and policy systems may grant trust based on an understated description, leading to unexpected network access, content transformation, and outbound publication.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The README instructs users to configure long-lived WeChat AppID/AppSecret credentials and notes that content will be uploaded to the WeChat draft box, but it does not warn about safe secret handling, storage, or the privacy implications of sending article content to a third-party platform. In an agent/skill context, this omission can lead users to expose secrets in shell history, shared environments, screenshots, or prompts, and to unknowingly transmit sensitive content to WeChat.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
95% confidence
Finding

The skill directs the agent to use shell/Python execution, WebFetch, file reads/writes, and publishing scripts, but it does not declare any tool scope or allowed-tools boundary. That creates an authorization gap where a runtime may permit broader tool use than users expect, increasing the chance of unintended command execution, network access, or data handling.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The manifest advertises 'No file conversion,' but the workflow explicitly runs a conversion script, rewrites HTML for compatibility, and may perform manual editing. Misrepresenting transformation behavior is dangerous because it obscures data modification, making users less able to assess integrity, compliance, and publication risk.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The manifest advertises 'No file conversion,' but the workflow explicitly runs a conversion script, rewrites HTML for compatibility, and may perform manual editing. Misrepresenting transformation behavior is dangerous because it obscures data modification, making users less able to assess integrity, compliance, and publication risk.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill lacks a clear up-front warning that it may fetch third-party URLs and publish content directly into a user's WeChat account. In a publishing skill, missing consent and risk disclosure is significant because it combines network retrieval, content processing, and an outbound side effect to an external account.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The documentation repeatedly assures users that no file conversion occurs, while later sections direct automated conversion and manual rewriting. Repeated inconsistent statements increase the likelihood that users misunderstand what data will be changed or sent, undermining informed consent and review before publication.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The manifest description and tags present the skill entirely in Chinese and target WeChat public account publishing, with no indication that users may choose another language or locale. This creates a natural-language policy concern because the skill appears to enforce a specific language context without documenting user choice or opt-in.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

Auto-installing Pillow changes the host environment and executes package installation logic that is unrelated to the core task of image compression. This is dangerous because package installation may run arbitrary setup/install hooks from fetched artifacts or trusted mirrors, and users invoking a local image helper may not expect network access or dependency mutation.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The script performs package installation without prior confirmation, so a routine file-processing action can unexpectedly launch a subprocess, access the network, and alter the Python environment. In an agent skill context, this is more dangerous because tools may run unattended or in privileged automation environments where silent package installation increases supply-chain and operational risk.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
91% confidence
Finding

The script automatically invokes pip in a subprocess when Pillow is missing, which causes code execution and environment modification as a side effect of processing an image. Even though the command is fixed and not shell-injected, installing packages at runtime expands the trust boundary to package indexes, local pip configuration, and network state, making a simple utility unexpectedly capable of executing additional code.

Content

Scanner excerpt · scripts/compress_image.py (reported line 33)May include surrounding context.

python
except ImportError:
    print("[INFO] Installing Pillow...")
    import subprocess
    subprocess.check_call([sys.executable, "-m", "pip", "install", "Pillow", "-q"])
    from PIL import Image

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
77% confidence
Finding

The README presents the skill description, examples, and user invocation phrases entirely in Chinese, which can amount to a fixed language assumption if the skill is intended for general use. There is no indication that users may interact in other languages or that Chinese is a deliberate opt-in or region-specific requirement.

Content

No source excerpt is available for this finding.