The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the official aliyun CLI using subprocess.run in scripts/_oss_client.py for identity verification. This implementation uses a structured argument list without a shell (shell=False), effectively preventing command injection vulnerabilities. The executed command is restricted to the non-mutating sts get-caller-identity operation.
[EXTERNAL_DOWNLOADS]: The skill interacts with official Alibaba Cloud service endpoints (e.g., *.aliyuncs.com) and fetches documentation index files and content from help.aliyun.com. These are trusted vendor resources provided by the skill's author (aliyun). The scripts/_doc_lookup.py module implements a strict host whitelist to ensure it only retrieves data from the official documentation domain.
[CREDENTIALS_SAFE]: The skill strictly follows least-privilege principles by prohibiting manual input or transmission of credentials. It relies on the standard Alibaba Cloud environment-based credential chain (e.g., ALIBABA_CLOUD_ACCESS_KEY_ID), which is a secure practice for cloud-native agent skills.
[SAFE]: All diagnostics are enforced as read-only. The skill contains explicit logic and documentation prohibiting the use of mutating APIs (like PutBucketCors). It provides configuration templates as text output for users to review and apply manually, maintaining human-in-the-loop oversight for all infrastructure changes.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input such as bucket names, origins, and error messages. It mitigates injection risks through robust sanitization, including regex validation for bucket names and normalization for origin strings. These inputs are used primarily for read-only cloud metadata queries and keyword matching against a local documentation index.

alibabacloud-oss-browser-upload-cors-diagnosis