wechat-styler

Pass

Audited by Gen Agent Trust Hub on Jul 27, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill invokes several local system commands to perform its duties. It uses opencli for interacting with the WeChat Official Account editor in the browser, sips on macOS for image optimization (resizing and format conversion), and the open command to present generated HTML files to the user. These executions are part of the core functionality and are handled with structured arguments.
  • [EXTERNAL_DOWNLOADS]: During the image optimization pipeline, the skill fetches remote images referenced in the Markdown articles using the standard fetch API. This is done to ensure images meet WeChat's size and format constraints before they are uploaded to the publishing platform.
  • [DATA_EXFILTRATION]: The skill's primary purpose is to transmit generated content (HTML and images) to the WeChat editor. While this involves sending data to an external service, it is the intended behavior of a publishing tool. The implementation includes a redactUrl utility in scripts/wechat-opencli.mjs that specifically masks sensitive token= parameters in URLs before they are written to reports or logs, mitigating accidental credential exposure.
  • [DYNAMIC_EXECUTION]: The skill uses opencli eval to execute JavaScript snippets within the context of the WeChat editor webpage. These scripts are used for DOM manipulation (injecting content) and status verification. The construction of these scripts in scripts/wechat-publish-core.mjs uses JSON.stringify and Base64 encoding to safely transport data, and the tool includes a dedicated validator (scripts/validate.mjs) that flags potentially malicious HTML tags like <script> or <iframe> in the source Markdown.
  • [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow where an AI agent is instructed to "intelligently rewrite" Markdown content to include structural components. This creates a surface where malicious instructions in a user-provided Markdown file could influence the agent's behavior. However, the risk is categorized as low because the downstream conversion and injection processes are deterministic and include safety validation steps.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 27, 2026, 04:55 PM
Security Audit — agent-trust-hub — wechat-styler