The Agent Skills Directory

Credential and Environment Auditing: The skill suggests using shell commands like grep to scan the local project for hardcoded secrets (e.g., api_key) and potentially unsafe environment variable usage. This is a defensive diagnostic measure intended to help users move secrets to secure providers.
IAM Policy Scoping Guidance: It provides templates and commands for auditing and restricting AWS IAM permissions, such as limiting Bedrock model access and scoping ECR repository permissions, following the principle of least privilege.
Administrative Tool Integration: The skill interacts with the aws and agentcore CLI tools to retrieve project status and configure service quotas. These operations are standard for AWS environment management.
Indirect Prompt Injection Surface: The skill ingests data from local files like agentcore/agentcore.json to generate its checklist. While this is necessary for its functionality, it represents a surface where local project metadata can influence the agent's responses.

agents-harden