The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous commands and scripts for managing AWS infrastructure via the AWS CLI and boto3. These operations, including IAM role creation and resource provisioning, are strictly aligned with the skill's primary purpose as a cloud development toolkit.
[EXTERNAL_DOWNLOADS]: The skill references and fetches documentation from the official 'aws-samples' GitHub repository, which is a trusted and well-known source.
[PROMPT_INJECTION]: The skill manages untrusted data through the 'x402_fetch' tool and Knowledge Base ingestion, which introduces a surface for indirect prompt injection. The skill includes built-in mitigations to manage this risk.
Ingestion points: The 'x402_fetch' tool retrieves content from external URLs (references/agentcore-payments-wiring.md), and Knowledge Bases ingest data from S3, SharePoint, and other sources (references/knowledge-bases-setup.md).
Boundary markers: The skill explicitly instructs the agent to treat all agent-generated parameters and external data as untrusted input.
Capability inventory: The skill utilizes AWS CLI commands, boto3 API calls for cloud management, and the 'httpx' library for network requests.
Sanitization: The 'x402_fetch' implementation includes a validation function that enforces HTTPS and blocks access to private or internal IP ranges to prevent SSRF.
[SAFE]: The skill adheres to security best practices by recommending IAM roles over long-lived credentials and instructing users to store sensitive keys in AWS Secrets Manager.

amazon-bedrock