Hi Ary, I maintain Skillgate, which checks a project's definition of done independently of the coding agent.
PlanQueue's prompt queue and human checkpoints look like a useful fit for a small optional completion-check recipe:
- Run a queued prompt.
- Run the project's configured completion check when the turn settles.
- Advance only when it passes; otherwise pause with the failed checks visible.
The default queue behavior would stay as it is. I would start with one documented example and a generic command/exit-code boundary, without making Skillgate a required dependency. Its existing demo shows the check failing, blocking a commit, then passing after the fix.
This is adjacent to #2's command-execution request, but the question here is whether a check should control progression to the next item. I haven't built or verified a PlanQueue integration yet.
Would a small companion recipe fit the project's scope? If so, I'm happy to contribute an example. Also curious which of my tools first caught your interest.
Hi Ary, I maintain Skillgate, which checks a project's definition of done independently of the coding agent.
PlanQueue's prompt queue and human checkpoints look like a useful fit for a small optional completion-check recipe:
The default queue behavior would stay as it is. I would start with one documented example and a generic command/exit-code boundary, without making Skillgate a required dependency. Its existing demo shows the check failing, blocking a commit, then passing after the fix.
This is adjacent to #2's command-execution request, but the question here is whether a check should control progression to the next item. I haven't built or verified a PlanQueue integration yet.
Would a small companion recipe fit the project's scope? If so, I'm happy to contribute an example. Also curious which of my tools first caught your interest.