The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell commands using the aliyun CLI (e.g., aliyun sts get-caller-identity, aliyun sls get-logs). These operations are performed using argument lists in subprocess.run (avoiding shell=True) and are limited to the official vendor toolset for identity verification and log retrieval.
[EXTERNAL_DOWNLOADS]: The _doc_lookup.py script fetches documentation from help.aliyun.com. This is a well-known service belonging to the vendor (Alibaba Cloud) and is used solely for read-only documentation verification, with host-level validation to prevent redirection to untrusted domains.
[CREDENTIALS_SAFE]: The skill implements secure credential management by exclusively using the Alibaba Cloud default credential chain (environment variables like ALIBABA_CLOUD_ACCESS_KEY_ID). It explicitly prohibits the manual passing or printing of secrets and utilizes the oss2 SDK's standard providers.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data in the form of OSS access logs. However, the data is ingested via structured Python scripts that perform type coercion and statistical analysis (_log_evidence.py), treating the content as data rather than instructions. This significantly limits the risk of the agent being influenced by malicious content embedded in log fields like User-Agents or Referers.

alibabacloud-oss-security-incident-forensics