The Agent Skills Directory

[SAFE]: The skill provides documentation and CLI command templates for AWS Cognito services (User Pools and Identity Pools). It serves as an instructional resource for developers and AI agents to configure authentication workflows.
[DATA_EXFILTRATION]: The instructions explicitly warn against data exfiltration and credential exposure. For example, it provides detailed patterns for using umask 077 and temporary files to avoid exposing social identity provider secrets in shell history or process lists. It also advises against storing sensitive tokens in localStorage due to XSS risks.
[EXTERNAL_DOWNLOADS]: The skill references official AWS-maintained libraries, such as aws-amplify and aws-jwt-verify (from the trusted awslabs GitHub organization). These are well-known, trusted resources within the AWS ecosystem.
[COMMAND_EXECUTION]: The skill includes numerous AWS CLI command templates (aws cognito-idp, aws cognito-identity). These are used for administrative configuration of the user's own AWS environment and do not involve suspicious execution of external or untrusted scripts.
[INDIRECT_PROMPT_INJECTION]: The skill documents how to implement Lambda triggers which process external user data. It proactively addresses security by instructing the agent to 'fail closed' on authentication decisions and to sanitize PII in logs, mitigating the risk of instructions embedded in user data influencing system behavior.

aws-auth