The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes git commands through subprocess.run to synchronize deployment configurations. It effectively mitigates command injection risks by strictly sanitizing deployment names using a restrictive regular expression and prevents the execution of malicious repository hooks by setting core.hooksPath=/dev/null.
[CREDENTIALS_SAFE]: The implementation ensures that sensitive authentication data, such as session cookies, is not leaked by using a dedicated redaction function to scrub secrets from all outputs and logs. Additionally, local configuration files containing resource scope information are protected with restricted filesystem permissions (0o600).
[EXTERNAL_DOWNLOADS]: The skill relies on official Alibaba Cloud SDK libraries as dependencies. All network communications are established with legitimate vendor endpoints on the aliyuncs.com domain for the purpose of job orchestration and metric retrieval.
[INDIRECT_PROMPT_INJECTION]: The skill manages the ingestion of external data, such as execution logs and platform metrics. It reduces susceptibility to indirect prompt injection by utilizing structured output templates and explicit instruction boundaries, ensuring that the agent does not treat untrusted log content as authoritative instructions.

alibabacloud-flink-sql-skill