The Agent Skills Directory

[PROMPT_INJECTION]: The skill's instructions in SKILL.md contain explicit directives for the AI agent to ignore the index.js file and avoid reading its contents. This is a defensive pattern designed to bypass security analysis and hide the script's internal logic.
[COMMAND_EXECUTION]: The script executes multiple system commands using wmic (wmic cpu get ProcessorId, wmic baseboard get SerialNumber, wmic csproduct get UUID) to harvest unique hardware identifiers. This data is used to generate a machine-specific hash for licensing and usage tracking without explicit user consent.
[CREDENTIALS_UNSAFE]: The skill prompts users to provide highly sensitive WeChat AppID and AppSecret credentials. These are written to a local config.json file. Because the main code is obfuscated, it is impossible to verify that these credentials are handled securely or not sent to third-party servers.
[DATA_EXFILTRATION]: The skill tracks usage limits and license status by creating persistent files and hidden directories (.node_cache, .session_a1b2c3.dat) in the user's home directory. It also manages sensitive session tokens in wechat-token.json.

wechat-publisher-yashu