The Agent Skills Directory

[COMMAND_EXECUTION]: The Python script scripts/kb.py utilizes subprocess.run to execute the git show command. This is used to read content from a specified repository for knowledge base retrieval. The script enforces strict validation on the commit SHA (hexadecimal only) and the file path (no parent directory references or absolute paths) to mitigate command injection risks.
[EXTERNAL_DOWNLOADS]: The skill interacts with the Aliyun Codeup API (openapi-rdc.aliyuncs.com) to search for and retrieve business and development rules. This interaction is performed neutrally for data retrieval purposes and utilizes authorization tokens managed through environment variables.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface by ingesting data from external Aliyun Codeup repositories and performing visual assertions on user-specified test URLs in assets/midscene/tests/ai.spec.ts. This risk is categorized as low because the data is processed as passive content, and the skill's execution capabilities are limited to sanitized, read-only repository commands.
[SAFE]: A dedicated integrity verification script, scripts/doctor.py, checks the SHA256 hashes of the skill's files against a manifest (package-manifest.json). This ensures that the skill components have not been tampered with and maintain their original trusted state.

oneos-work-verify