The Agent Skills Directory

[DATA_EXPOSURE]: The skill provides SQL queries to access crdb_internal.kv_store_status. This table contains metadata about the cluster's physical storage layout, including node_id, store_id, capacity, and current disk usage. While this provides visibility into internal cluster state, it is standard diagnostic data required for the skill's stated purpose of capacity planning.
[COMMAND_EXECUTION]: The skill provides complex SQL templates for execution against a CockroachDB cluster. These queries use advanced features like SHOW RANGES WITH DETAILS and JSON path extraction (span_stats->>'approximate_disk_bytes'). The instructions include explicit warnings about the CPU and memory overhead of these commands and provide guardrails such as the use of LIMIT and specific table targeting.
[INDIRECT_PROMPT_INJECTION]: The skill uses placeholders (e.g., your_table_name) in SQL queries.
Ingestion points: Table names are provided as input to SQL templates in SKILL.md and storage-calculations.md.
Boundary markers: Absent; the queries are provided as plain text templates.
Capability inventory: SQL execution via an established database connection.
Sanitization: Not applicable as these are documentation templates intended for an agent to populate and execute.
[PRIVILEGE_MANAGEMENT]: The skill explicitly documents its requirement for admin or ZONECONFIG privileges. It provides a standard method for users to verify their current grants (SHOW GRANTS ON SYSTEM) before attempting restricted operations, promoting best practices for role-based access control.

analyzing-schema-change-storage-risk