The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes external binaries (bdpan and baidupan-cli) using Python's subprocess.run. The implementation in scripts/run_with_env.py and scripts/scan_drive.py explicitly sets shell=False, which is a security best practice to prevent shell command injection.
[EXTERNAL_DOWNLOADS]: The skill requires external dependencies including a separate agent skill from GitHub (baidu-netdisk/bdpan-storage) and optionally a community-developed CLI tool. It provides clear instructions for manual installation and review of these components.
[CREDENTIALS_UNSAFE]: While the skill handles sensitive application keys and secrets, it implements a secure pattern by loading these values from a local configuration file into the process environment variables just before executing the CLI tools, minimizing the risk of exposure in logs or terminal history.
[DATA_EXFILTRATION]: The skill interacts with Baidu Netdisk APIs through official and recognized community tools. Analysis of the Python scripts confirms that data retrieved from the cloud is only processed locally for scanning or displayed to the user, with no hidden network exfiltration paths.
[PROMPT_INJECTION]: The skill incorporates safeguards against directory traversal in its scan_drive.py script by validating that remote paths do not contain .. or ~, reducing the risk of indirect manipulation via malicious file structures.

soia-pkm-baidupan