The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches documentation from help.aliyun.com to verify error codes and best practices. As this domain is an official service domain of the vendor (aliyun) and a well-known service, it is considered safe.
[COMMAND_EXECUTION]: The skill uses subprocess.run in scripts/_cli.py and scripts/_oss_client.py to invoke the aliyun CLI. These calls use argument-list formatting and disable stdin to prevent shell injection or interactive hangs. The operations are restricted to a whitelist of read-only actions.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from Alibaba Cloud SLS logs and processes it to diagnose errors. While this presents an attack surface, the risk is mitigated by structured parsing, strict field mapping, and the skill's own "honesty contract" which prevents the AI from fabricating conclusions when data is missing or ambiguous. The severity is categorized as low in accordance with the security posture rules.
[CREDENTIALS_UNSAFE]: The skill explicitly instructs the agent NOT to read, print, or store credentials from environment variables or local configuration files like ~/.aliyun/config.json. It relies on the system's default credential chain without manual interference.

alibabacloud-oss-access-log-trace-diagnosis