The Agent Skills Directory

[SAFE]: The skill implements significant safety measures for database operations. It defaults to read-only mode and requires explicit --write flags and validated credentials for any modifications. Security is reinforced via database-specific drivers (e.g., SQLITE_OPEN_READONLY for SQLite, read-only transactions for PostgreSQL).
[SAFE]: Credential management follows security best practices. The skill explicitly instructs users not to paste secrets into the LLM context, instead utilizing a --form flag to trigger native OS dialogs that keep passwords outside of the agent's visibility.
[SAFE]: Network access is restricted to established database drivers (PostgreSQL, MySQL, Snowflake, etc.) and is used exclusively for its primary purpose of database querying and schema exploration.

agent-sql