The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes Mini-Program source code and runtime diagnostic logs which could contain malicious instructions if the project being debugged is attacker-controlled.
Ingestion points: Ingests data from wechat_file.read_page, wechat_file.read_file, wechat_inspector.cdp, wechat_inspector.console, and wechat_automator.page_data.
Boundary markers: No explicit boundary markers or delimiters are described to separate ingested data from agent instructions.
Capability inventory: Comprehensive Mini-Program interaction including file reading, IDE CLI command execution, UI automation (tap/input), and arbitrary JavaScript evaluation (wechat_automator.evaluate).
Sanitization: No sanitization or filtering of ingested code or log content is mentioned.
[DYNAMIC_EXECUTION]: The wechat_automator tool includes an evaluate action that executes JavaScript source code (fn_source or expression) within the WeChat Mini-Program logic layer (AppService). This is a core functionality for automation and debugging, but involves the runtime execution of arbitrary strings.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of the wechat-devtools-mcp tool via uv. This package is a vendor-provided resource from the skill author (WaterTian) designed to bridge the MCP environment with the WeChat Developer Tools.

wechat-devtools