What problem does it solve? When your AgentCore agent returns wrong answers, throws errors, times out, or the CLI itself stops working, this Skill reads traces and logs directly to find the root cause instead of asking you to paste output manually. ## Core Features & Use Cases - Trace and log analysis: Automatically lists, downloads, and interprets recent traces and CloudWatch logs to pinpoint failing model calls, tool calls, and memory operations. - Symptom-based diagnosis playbooks: Covers model access denied, tool call failures, memory not persisting, 424 errors, maxVms quota exhaustion, streaming disconnects, missing CloudWatch logs, and stuck DELETING runtimes. - Environment doctor: Checks CLI version, Node.js, uv, AWS credentials, Bedrock model access, IAM permissions, and Docker with exact fix commands for each failure. - Use Case: Your deployed agent returns "model access denied" — the Skill checks Bedrock model access, verifies the execution role has bedrock:InvokeModel, and explains cross-region inference profile requirements. ## Quick Start Ask the agent to debug why your AgentCore agent is failing by running the agents-debug skill and describing the symptom.
