md2wechat

Warn

Audited by Gen Agent Trust Hub on Jul 15, 2026

Risk Level: MEDIUMCREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill defines installation procedures for the md2wechat CLI tool from a third-party GitHub repository (geekjourneyx/md2wechat-skill) using brew and go install. These sources are not from a verified or trusted vendor.
  • [CREDENTIALS_UNSAFE]: The skill requires sensitive WeChat credentials (WECHAT_APPID and WECHAT_SECRET) to perform draft creation and publishing actions. Handling these secrets in the agent's environment presents a risk of exposure.
  • [PROMPT_INJECTION]: The instructions include specific commands to remove AI writing traces and humanize content. These directives are intended to bypass AI detection mechanisms and content safety filters.
  • [DATA_EXFILTRATION]: The CLI tool supports a configurable API base URL via MD2WECHAT_BASE_URL. If this configuration is pointed to an unverified or malicious endpoint, it could be used to intercept sensitive article content or API credentials.
  • [COMMAND_EXECUTION]: The skill relies on executing a locally installed binary and external Node.js scripts with file system and network access. It processes untrusted input from Markdown files and remote URLs. 1. Ingestion points: Local Markdown files and remote image URLs. 2. Boundary markers: Absent. 3. Capability inventory: File system writes, network uploads to WeChat APIs, and remote image generation. 4. Sanitization: Not specified for the processed content.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 15, 2026, 07:56 AM
Security Audit — agent-trust-hub — md2wechat