The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for executing administrative AWS CLI commands to manage database infrastructure, including restoring instances from snapshots and modifying network security settings.
Evidence: SKILL.md contains commands such as aws rds restore-db-instance-from-db-snapshot and aws rds wait db-instance-available to facilitate infrastructure management.
[DATA_EXFILTRATION]: The workflow's primary goal is to export database content into a local archive file, which involves processing sensitive data.
Evidence: The use of pg_dump in Step 5 of SKILL.md to create a DUMP_FILE from the restored instance.
[INDIRECT_PROMPT_INJECTION]: The skill identifies an attack surface for indirect prompt injection as it interpolates user-supplied identifiers into shell commands.
Ingestion points: Variable declarations for PROFILE, REGION, SOURCE_DB, and SNAPSHOT_ID in SKILL.md.
Boundary markers: The instructions lack explicit shell escaping or validation for these interpolated variables.
Capability inventory: The agent is granted capabilities to execute shell commands, manage AWS RDS resources, and write files to the local system.
Sanitization: No programmatic sanitization is performed on the user-provided variables within the instructions.

rds-snapshot-dump