Improve troubleshooting execution steering guide#161
Conversation
Expand trigger keywords in POWER.md for better routing to the troubleshooting guide. Rewrite troubleshooting-executions.md with clearer diagnostic steps, ARN-based workflow, console link generation, structured output format, and additional usage examples.
02fb759 to
6260c96
Compare
|
Testing prompts |
|
We'll need to backport this to https://github.com/awslabs/agent-plugins/blob/main/plugins/aws-serverless/skills/aws-lambda-durable-functions/references/troubleshooting-executions.md as well. |
|
|
||
| 3. If the command fails: | ||
| - Execution not found: Tell the user the execution ID may be incorrect or the execution may have been purged. Ask them to verify the ARN. | ||
| - Permissions/network error: Suggest checking IAM permissions (lambda:GetDurableExecutionHistory) |
There was a problem hiding this comment.
suggestion:
Permissions/network error: check that your caller identity has ambda:GetDurableExecutionHistory on the function ARN. This is not part of AWSLambdaBasicDurableExecutionRolePolicy."
There was a problem hiding this comment.
This is a good point - general note, do we clarify anywhere in the power about the permission model? I think it might be useful to have some context about permissions you need within the lambda loop (e.g. checkpoint, get state - included in AWSLambdaBasicDurableExecutionRolePolicy), and permissions outside of the loop (like this one).
There was a problem hiding this comment.
I'll add "Permissions/network error: check that your caller identity has ambda:GetDurableExecutionHistory on the function ARN."
don't think we need to mention "AWSLambdaBasicDurableExecutionRolePolicy" since the user could be signed in with any role.
Add steps to fetch the log group and query logs filtered by invocation ID and execution name when execution history alone is insufficient to determine root cause.
Add step 0 to resolve function alias to version and list executions when user provides function name instead of full ARN. Also add STOPPED status and improve permissions error guidance.
|
Test prompt with cloudwatch logs |
Backport changes from aws/aws-durable-execution-docs#161 to the agent-plugins repo. Improvements include: - ARN-based workflow instead of separate function name + execution ID - Alias resolution and list executions step - Console link generation for visual debugging - Structured output format with clear diagnosis summaries - CloudWatch Logs querying when execution history is insufficient - Additional usage examples covering failure scenarios - Better error handling guidance (execution not found, permissions) - Expanded trigger keywords in SKILL.md for better routing Closes aws/aws-durable-execution-docs#162
Backport changes from aws/aws-durable-execution-docs#161 to the agent-plugins repo. Improvements include: - ARN-based workflow instead of separate function name + execution ID - Alias resolution and list executions step - Console link generation for visual debugging - Structured output format with clear diagnosis summaries - CloudWatch Logs querying when execution history is insufficient - Additional usage examples covering failure scenarios - Better error handling guidance (execution not found, permissions) - Expanded trigger keywords in SKILL.md for better routing Closes aws/aws-durable-execution-docs#162
Backport changes from aws/aws-durable-execution-docs#161 to the agent-plugins repo. Improvements include: - ARN-based workflow instead of separate function name + execution ID - Alias resolution and list executions step - Console link generation for visual debugging - Structured output format with clear diagnosis summaries - CloudWatch Logs querying when execution history is insufficient - Additional usage examples covering failure scenarios - Better error handling guidance (execution not found, permissions) - Expanded trigger keywords in SKILL.md for better routing Closes aws/aws-durable-execution-docs#162
Summary
Improves the troubleshooting-executions steering file with:
Testing
Verified changes are limited to the two steering files. No code changes.
Attached the prompts used to test this. check this comment