The Agent Skills Directory

[PROMPT_INJECTION]: No instructions were found that attempt to override agent behavior or bypass safety constraints. The content is purely instructional and technical.
[DATA_EXFILTRATION]: The skill does not contain any network operations or commands to access sensitive system files. All SQL examples use generic placeholders and internal database constructs.
[CREDENTIALS_UNSAFE]: No hardcoded credentials or secrets were found. The security model reference uses standard placeholders for environment-supplied passwords in deployment templates.
[REMOTE_CODE_EXECUTION]: The skill does not reference or download any external scripts, binaries, or packages.
[COMMAND_EXECUTION]: Usage of SQL commands like EXEC is restricted to legitimate database administration tasks within templates, such as dropping constraints with dynamically determined names during migrations.
[SAFE]: The skill promotes high-security standards, including row-level security through views and restricting mutations to stored procedures, which aligns with the principle of least privilege.

sql-writing-guidelines