PRP Worklist
Incubating. Deliberately underspecified. Make your own judgements, then say what you chose and why — the good choices get recorded and prescribed later. Do not wait for a spec that does not exist yet.
Render the open issues and PRs of a repository as an artifact the maintainer can look at, and offer it. Not a list of everything — a view of what to take next.
The job
A tracker sorts by number and shouts every label equally loudly. That is the tool's view, not the maintainer's. Produce the maintainer's view.
Read the tracker with gh. Then decide what is worth showing and how to show it.
What this maintainer cares about
Considerations, not fields to emit. Weigh them, drop the ones that do not apply, add what the repository makes obvious:
- Blocked on a decision, or takeable now. The distinction he asks for most.
- Stale. An issue whose premise has moved — the code it describes was deleted, the blocker it waits on closed. Staleness should be visible, not discovered on starting the work.
- Near-duplicates. Two issues that are one thing. Cheap to spot in a list, expensive to find after both are filed.
- His, or someone else's. Different obligations as a maintainer.
- Blast radius rather than severity. What breaks, and for whom, if this is left.
Shape
Yours to decide. Markdown or HTML, grouped or ranked, dense or spacious — whatever makes the answer to "what do I take next" obvious in a glance.
Then offer it rather than pasting it: write the file to this project's ~/.prp/<key>/ store
and print a link the operator can open. If a helm-canvas skill is available, follow it.
Report your choices
End with a short note: what you grouped by, what you left out, what you could not tell from the tracker alone. That note is the point of this skill being loose — it is how the shape gets decided by evidence rather than up front.
If something here got in your way, say that too.
