The Agent Skills Directory

[SAFE]: The skill facilitates the use of official AWS tools, including the @aws/agentcore NPM package and bedrock-agentcore Python library, for managing agent lifecycles on Amazon Web Services.
[EXTERNAL_DOWNLOADS]: The skill references standard development dependencies and verified resources from well-known providers, including AWS repositories on GitHub, the Astral UV image from GitHub Container Registry, and official AWS documentation. These downloads are required for the stated purpose of building and deploying agents.
[SAFE]: Security best practices are encouraged, such as using .env.local for local secret management and leveraging standard AWS IAM roles for execution permissions.
[PROMPT_INJECTION]: The skill provides templates that ingest untrusted user input via a prompt field in the invoke entrypoints. While these ingestion points lack explicit boundary markers or sanitization in the minimal code snippets, this is typical for a foundation harness and represents a standard surface for indirect prompt injection that developers should address during implementation.

aws-harness