The Agent Skills Directory

[PROMPT_INJECTION]: No prompt injection or behavior override patterns were detected. The instructions follow standard educational and advisory language.
[DATA_EXFILTRATION]: No data exfiltration or unauthorized network access patterns were detected. The skill correctly recommends security best practices, such as using IAM Roles via OIDC for GitHub Actions to avoid long-lived secrets.
[EXTERNAL_DOWNLOADS]: The skill references official GitHub Actions from the 'actions' and 'aws-actions' organizations. These are well-known, trusted sources for CI/CD workflows and do not escalate the security risk.
[COMMAND_EXECUTION]: Provides standard AWS CLI and JSON configuration examples for managing ECS services. These examples are informative and aligned with the skill's primary purpose of deployment orchestration.
[CREDENTIALS_UNSAFE]: No hardcoded credentials, API keys, or private keys were found. The skill explicitly guides users toward secure credential management using AWS OIDC.
[REMOTE_CODE_EXECUTION]: No remote code execution patterns were detected. References to GitHub Actions are for user-configured CI/CD pipelines, not for runtime execution by the agent itself.
[INDIRECT_PROMPT_INJECTION]: The skill provides documentation on processing ECR scan results and task definition metadata. It defines clear boundaries for data ingestion in the context of deployment pipelines without introducing exploitable surfaces.

ecs-devops