During implementation, defer a suggestion from the chat for later review (implementing a new command /issue) #641
Description
Whilst developing a new project and interacting with the AI chat during the /implement phase I find that it makes suggestions for changes and improvements. When this happens, you have two choices, the first is to follow the suggestions immediately, the second is to ignore the suggestion for now, maybe making a mental note of it, or manually adding some details to the task list.
I propose adding a new command /issue that takes the relevant part of the current conversation and documenting it in a new "issue" file. A list of issues can then be added to the tasks.md file under a heading of "issues".
The agent can then at a later time review and resolve the issue.
I have already developed and tested a prompt for this, and added my own command to the .github/prompts folder. I have tested it in vscode.
issues-template.md
issue.prompt.md
Example of issues section of tasks.md
## Issues
- [ ] I001 Dashboard metrics broadcast returns 401 (unauthenticated) – see `issue_I001.md`
- Dashboard background service (`MetricsBroadcastService`) calls `/dev/metrics/monitoring` without `Authorization` or `X-Tenant-ID` headers. API inline auth/tenant middleware blocks request before endpoint executes. No `metrics` SignalR events reach clients (only heartbeat + keep-alives). Tracked separately for later resolution.
Example of issue_I001.md file:
issue_I001.md
I am ready to contribute this, but based on your guidelines for contribution, I am opening an issue regarding the suggestion for a discussion regarding the suggested new feature.