The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the aliyun CLI using subprocess.run with argument lists in scripts/origin_config_diagnosis.py and scripts/_oss_client.py. This is used exclusively for read-only operations like sts get-caller-identity and cdn describe-cdn-domain-detail. Command injection is prevented by avoiding shell execution and validating the --domain input against a strict regular expression.
[EXTERNAL_DOWNLOADS]: The _doc_lookup.py script fetches content from help.aliyun.com to verify configuration logic against official documentation. This is a well-known vendor service, and the script implements a host whitelist and strict URL validation to ensure only trusted content is retrieved.
[INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection via the --question parameter used for documentation lookup. Evidence chain: 1) Ingestion points: scripts/origin_config_diagnosis.py. 2) Boundary markers: Present in SKILL.md (Mandatory entry-point enforcement and Execution Principle) which direct the agent to rely on structured script findings. 3) Capability inventory: subprocess.run and urllib.request.urlopen (limited to GET). 4) Sanitization: _doc_lookup.py implements _clean_excerpt to strip markdown and limit text length. The primary diagnostic logic remains isolated from this input.
[CREDENTIALS_UNSAFE]: The skill follows secure credential management by utilizing the Alibaba Cloud default credential chain (environment variables and local CLI configuration). It explicitly prohibits hardcoding or passing AccessKeys directly in the instructions or script arguments.

alibabacloud-oss-cdn-origin-config-diagnosis