The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/publish.py utilizes subprocess.run to call a local helper script (render.py). This execution is performed using the current Python interpreter (sys.executable) and targets a file within the skill's own directory, which is a secure and standard way to implement modular logic.
[CREDENTIALS_UNSAFE]: The skill handles WeChat API credentials (WECHAT_APP_ID, WECHAT_APP_SECRET). It correctly instructs the user to store these in a private local directory (~/.config/soia-skills/...) and provides a dedicated loader (scripts/publish_env.py) to inject them into the environment at runtime. This avoids hardcoding secrets and follows standard secret management practices.
[DATA_EXFILTRATION]: The skill reads configuration files and user articles from the local filesystem. These are used solely for the intended purpose of communicating with official WeChat services (mp.weixin.qq.com). No evidence of unauthorized data transmission or exfiltration to third-party servers was found.
[REMOTE_CODE_EXECUTION]: All scripts included in the skill are local and perform deterministic processing of data. There are no patterns involving the download and execution of remote code or the use of unsafe functions like eval() or exec() on untrusted data.
[EXTERNAL_DOWNLOADS]: The skill documentation references installation via the soia-team vendor's own marketplace and standard plugin tools. These represent legitimate deployment channels for the skill author and do not involve suspicious third-party repositories.

soia-media-publish-wechat-draft