The Agent Skills Directory

[SAFE]: The skill defines a content trust boundary for user-supplied input such as SQL statements and incident titles, explicitly instructing the agent to ignore any instructions embedded within that data to prevent prompt injection.- [SAFE]: Input validation is enforced for CLI arguments, with specific instructions to reject shell metacharacters such as backticks, pipes, and semicolons, effectively mitigating command injection risks.- [SAFE]: The skill utilizes temporary files to pass large GraphQL queries and variables to the CLI, which is a recommended practice to avoid shell word-splitting issues and command-line length limitations.- [SAFE]: Security guardrails are implemented for custom SQL assertions, requiring the agent to warn the user and refuse to execute destructive commands like DROP, TRUNCATE, or DELETE.- [SAFE]: The skill operates using the official datahub CLI and targets vendor-controlled resources, ensuring behavior is consistent with its intended administrative purpose.

datahub-quality