The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests potentially untrusted data from endpoint configurations and system logs. Ingestion points: The skill retrieves external data using endpoint-get and execute-sql for query logs. Boundary markers: The instructions explicitly require the agent to get user sign-off and manually verify the semantic equivalence of rewrites before application. Capability inventory: The skill utilizes execute-sql for data retrieval and endpoint-update for modifying endpoint logic. Sanitization: The workflow relies on manual user verification and comparison of query results rather than automated filtering.
[COMMAND_EXECUTION]: The skill uses execute-sql to query the query_log table for auditing performance.
[DYNAMIC_EXECUTION]: The skill generates SQL query rewrites via a specialized suggestion tool and executes them using execute-sql for verification.

diagnosing-endpoint-performance