The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to analyze database system views, query logs, and execution plans (e.g., pg_stat_statements, EXPLAIN output). These sources can contain arbitrary strings from external queries, creating a surface where malicious instructions embedded in a query could influence the agent's analysis.
Ingestion points: pg_stat_statements, pg_stat_activity, performance_schema, and EXPLAIN output referenced in SKILL.md and references/monitoring-analysis.md.
Boundary markers: None defined in the skill instructions to delimit or ignore instructions within database metadata.
Capability inventory: The skill generates SQL for DDL (CREATE INDEX), DML, and system configuration (ALTER SYSTEM, SET GLOBAL).
Sanitization: No specific sanitization or filtering logic is provided for the data read from database system views.
[COMMAND_EXECUTION]: The skill provides templates for high-privilege administrative SQL commands.
Reference files such as references/postgresql-tuning.md and references/mysql-tuning.md include commands like ALTER SYSTEM SET and SET GLOBAL. These operations modify global server configurations and typically require superuser or administrative database privileges.
[NO_CODE]: The skill consists entirely of Markdown documentation and SQL templates. It does not include any scripts (Python, JavaScript, Shell) or executable binaries.

database-optimizer