The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides SQL queries for database administration that include the ability to terminate backends (pg_terminate_backend) and cancel active queries (pg_cancel_backend). The documentation includes explicit safety rules requiring human confirmation before executing any destructive operations.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the processing of database activity by reading system views such as pg_stat_statements and pg_stat_activity, which contain query text and metadata potentially controlled by external database clients.
Ingestion points: Data enters the agent's context through queries in MONITORING.md that read from PostgreSQL system statistics views.
Boundary markers: No specific delimiters or warnings regarding embedded instructions are provided for the output of these monitoring queries.
Capability inventory: The skill includes the capability to execute administrative SQL commands on the target database.
Sanitization: There is no documented validation or sanitization of the content retrieved from database statistics views.

postgres