The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on running a local script, scripts/sql_query_helper.py, which is executed with user-supplied arguments to manage database schema extraction and query analysis.
[DATA_EXFILTRATION]: The skill accepts an arbitrary filesystem path via the --db-path argument for SQLite databases. This allows the agent to read any SQLite file it has permissions for, which could include sensitive application databases, browser history, or credential stores.
[EXTERNAL_DOWNLOADS]: The documentation indicates that the psycopg2-binary package is required for PostgreSQL support, which must be retrieved from a public package registry.
[PROMPT_INJECTION]: The skill ingests schema definitions and sample data from external databases, presenting a surface for indirect prompt injection.
Ingestion points: Table names, column types, and sample data rows are retrieved by scripts/sql_query_helper.py and included in the agent's context.
Boundary markers: Data extracted from the database is returned without specialized delimiters or warnings to the agent to disregard instructions within the data.
Capability inventory: The helper script can read files from the local filesystem and establish network connections to PostgreSQL servers.
Sanitization: While the script includes a keyword filter to prevent database write operations, it does not sanitize or escape the content retrieved from the database before it is presented to the agent.

sql-insight