The Agent Skills Directory

Indirect Prompt Injection Surface: The skill describes the use of aws ssm send-command for fleet operations, which provides a surface for indirect prompt injection. External or user-supplied input could potentially influence the execution of shell commands on managed nodes.
Ingestion points: External input built into --targets tag values or --parameters commands (referenced in references/systems-manager.md).
Boundary markers: The skill does not define structural delimiters for the input but provides explicit warnings to validate and escape it.
Capability inventory: The skill utilizes aws ssm send-command to perform shell execution across an EC2 fleet (references/systems-manager.md).
Sanitization: The instructions include a specific remediation note advising users to "validate and escape" input before use to prevent command-injection vectors.
Command Execution Patterns: The skill provides various shell commands for managing AWS infrastructure. While these involve command-line execution, they are standard administrative operations. The skill promotes security best practices, such as discouraging the placement of secrets in user data and recommending the use of encrypted Session Manager logs to maintain an audit trail and protect sensitive data.

aws-compute