The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill includes instructions to download the tosutil binary from the official vendor infrastructure (volces.com). As these downloads target the author's own verified domain for the purpose of tool installation, the activity is consistent with the skill's primary function.
[COMMAND_EXECUTION]: The skill executes the tosutil CLI using the subprocess module in Python. Security is maintained by passing arguments as a list and using shlex.quote for command previews, which prevents shell-based command injection from user-supplied inputs.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external data such as cloud URLs (tos://) and local file paths which are interpolated into commands. However, the risk is mitigated by explicit boundary checks, a required --run flag for execution, and the use of list-based subprocess calls that prevent inputs from being interpreted as commands.
[CREDENTIALS_UNSAFE]: While the skill processes sensitive cloud credentials (AK/SK/STS Tokens), it implements best-practice security controls including the redact_argv and mask_sensitive_text functions to ensure secrets are never leaked in logs, command previews, or standard output.

volcengine-tosutil