The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the user to execute various agentcore CLI commands to configure the development environment, manage credentials, and deploy gateway targets (e.g., agentcore add gateway-target, agentcore add credential).
[EXTERNAL_DOWNLOADS]: The documentation references the agentcore update command, which is a standard mechanism for maintaining the CLI tool's version.
[DATA_EXFILTRATION]: The skill facilitates the connection of agents to external APIs and services. It provides specific guidance on using the AgentCore Gateway to ensure that sensitive credentials remain at the edge and are never exposed to the agent's logic or logs.
[PROMPT_INJECTION]: The skill establishes a surface for Indirect Prompt Injection by enabling the ingestion of data from third-party MCP servers and OpenAPI-compliant APIs.
Ingestion points: External MCP servers, Lambda functions, and OpenAPI schemas referenced in SKILL.md.
Boundary markers: The skill does not explicitly define prompt boundary markers but recommends the use of the AgentCore Gateway for centralized policy enforcement.
Capability inventory: Uses httpx and the mcp library for tool invocation and network communication.
Sanitization: Relies on the Cedar policy engine to validate and authorize tool calls based on input parameters (e.g., restricting refund amounts).

agents-connect