The Agent Skills Directory

[COMMAND_EXECUTION]: The skill frequently directs the agent to execute database-native CLI tools (e.g., psql, mysql, mongosh, and sqlcmd) to verify cluster health, analyze query plans (EXPLAIN), and manage configurations. These operations occur across several reference files including SKILL.md and build-patterns.md.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from users, such as SQL queries, database schemas, and migration logic. This data is then used in tool calls and diagnostic commands. For example, in Step 4 of SKILL.md and in build-patterns.md, user-provided queries are analyzed using EXPLAIN. The skill mitigates this risk through mandatory 'AI Self-Checks', production checklists, and explicit warnings against performing destructive operations without direct user confirmation.
[DYNAMIC_EXECUTION]: The skill provides numerous bash script templates for complex operations like continuous backups, data backfilling, and restore verification (e.g., in references/backup-patterns.md and references/migration-patterns.md). These scripts are intended to be populated with runtime variables and executed to maintain production databases. The skill emphasizes best practices like parameterization and the use of protected configuration files (e.g., pgpass, MONGODB_BACKUP_CONFIG) rather than passing credentials via command-line arguments.

databases