The Agent Skills Directory

[SAFE]: The skill is designed for read-only diagnosis and only issues metadata queries (such as GetBucketInfo and GetBucketReferer) to analyze OSS configurations. It contains no mutating API calls.
[COMMAND_EXECUTION]: The skill uses the official aliyun CLI for identity verification. Analysis of the implementation shows that commands are executed using argument-list subprocess calls without shell expansion, which prevents command injection vulnerabilities.
[EXTERNAL_DOWNLOADS]: The skill fetches documentation excerpts from the official Alibaba Cloud help center. It implements a URL whitelist that restricts network requests to the help.aliyun.com domain, ensuring that only trusted content is retrieved.
[INDIRECT_PROMPT_INJECTION]: The skill processes user questions to verify them against official documentation.
Ingestion points: User input provided via the --question CLI argument in the main diagnosis script.
Boundary markers: Findings are presented within a structured JSON report and marked with specific status lines (STATUS: OK/DEGRADED) to separate them from general agent reasoning.
Capability inventory: Restricted to read-only OSS API calls and identity checks via the aliyun CLI.
Sanitization: The skill cleans Markdown syntax and truncates external documentation excerpts before they are presented to the user.

alibabacloud-oss-direct-access-link-diagnosis