The Agent Skills Directory

[PROMPT_INJECTION]: The skill includes defensive instructions to protect against indirect prompt injection by advising the agent to ignore any instructions found in user-supplied values like assertion descriptions or SQL statements. It establishes clear trust boundaries between system instructions and untrusted data.
[COMMAND_EXECUTION]: The skill safely utilizes the datahub CLI through the Bash tool. It implements security measures such as rejecting shell metacharacters in arguments and using temporary files for complex GraphQL queries to prevent command injection and environment-specific errors.
[SAFE]: The skill follows security best practices by requiring mandatory user approval for all write operations, verifying the user's deployment tier before proposing restricted actions, and using structured data for all data interactions.

datahub-quality