The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell commands for connectivity and infrastructure management. Evidence: The psql fallback workflow in SKILL.md uses aws dsql generate-db-connect-admin-auth-token and psql to execute plans. Evidence: SKILL.md identifies management scripts in the scripts/ directory for cluster lifecycle and bulk data operations.
[PROMPT_INJECTION]: The skill processes externally-sourced SQL data, creating an indirect prompt injection surface. This is mitigated by structured workflows and validation tools. Ingestion points: User-provided queries and external SQL source files (migrations, pg_dump). Boundary markers: Instructions mandate dsql_lint validation and safe_query.build() interpolation. Capability inventory: Database DDL/DML execution via MCP tools and shell execution via scripts. Sanitization: Use of dsql_lint for compatibility checking and safe_query.py for input validation and escaping.
[EXTERNAL_DOWNLOADS]: The skill references official documentation and code samples from trusted AWS domains. Evidence: Links to docs.aws.amazon.com and official sample repositories at github.com/aws-samples/aurora-dsql-samples.

aurora dsql