would be nice to also trigger at Stop and SubagentStop;
Usecases:
- this way claudecode can reflect opn a successful task and if criterias are met then propose creating a skill for that.
- Use the Stop and Subagent Stop to reflect on a skill used and optimize it.