The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run to execute the official wecom-cli and an optional user-provided helper script (WECOM_UPLOAD_HELPER). These executions are performed using list-based arguments in scripts/create_smartpage.py and scripts/doctor.py, which prevents shell command injection. The commands are limited to diagnostic checks and authorized WeCom operations.- [EXTERNAL_DOWNLOADS]: The skill instructions reference the official @wecom/cli package from the WecomTeam GitHub organization. This is a well-known and trusted service for interacting with the WeCom platform.- [DATA_EXFILTRATION]: The skill implements strong privacy boundaries. While it reads local Markdown files to upload content to WeCom, it explicitly avoids reading or outputting encrypted credentials from the configuration directory (~/.config/wecom/). Internal identifiers and upload receipts are stored locally with restricted file permissions (0600/0700) to prevent unauthorized access.- [PROMPT_INJECTION]: The skill processes content from local Markdown files (ingestion point in scripts/create_smartpage.py), which represents an indirect prompt injection surface. The skill mitigates this by requiring explicit user confirmation for all non-read operations (boundary marker in SKILL.md) and using JSON serialization for CLI arguments (sanitization). Cross-script capabilities are limited to authorized WeCom API calls via the official CLI.

yichen-wecom-operations