Starter kits

ticket-to-pr-ready — the reproduce-and-fix loop

Takes a ticket, bug report, or complaint and turns it into a review-ready patch: reproduce, prove the root cause, make the smallest credible fix,…

Takes a ticket, bug report, or complaint and turns it into a review-ready patch: reproduce, prove the root cause, make the smallest credible fix, re-verify. Says so honestly if it can't reproduce after two attempts. Never folds in an unrelated refactor.

Category: D→B · Issue intake → PR · Heartbeat: event (per ticket) · Level: L1 (report-only, drafts-not-opens-PR) · Cost: High

Source: Loop #16, "The ticket-to-PR-ready loop," from Forward Future's Loop Library, original prompt by Hiten Shah.

Quickstart

1. Read LOOP.md — the six parts, the prompt, the limits, the three stops,
   and the source's verbatim prompt.
2. ticket-to-pr-ready-state.md is your spine — already committed with an
   empty per-ticket register.
3. Claude Code: wire a ticket/issue-creation event trigger at
   .claude/skills/loop-task/SKILL.md. OpenCode: rename
   opencode.json.example → opencode.json, add the AGENTS.md line, wire
   skills/loop-task.md into the Actions event.
4. Rehearse once, by hand, on a real ticket before the trigger ever fires:
   "Run the ticket-to-pr-ready skill now on ticket #N, exactly as the event
   would."
5. Grade the rehearsal with the loop-verifier agent — it fails any patch
   that touches a file the fix didn't need, no matter how good the fix
   otherwise is. PASS → arm the trigger. FAIL → fix the skill, never widen
   the patch.

What's in this kit

File	Role
`LOOP.md`	the definition — six parts, prompt, limits, ownership, three stops
`ticket-to-pr-ready-state.md`	the spine — per-ticket reproduction evidence, root cause, fix, verification
`loop-budget.md`	caps: 20 runs, 300k tokens, 2 reproduction attempts/ticket
`loop-constraints.md`	the constitution — never folds in an unrelated refactor, never opens a PR at L1
`loop-run-log.md`	append-only, one line per beat
`.claude/skills/loop-task/SKILL.md`	the procedure (Claude Code)
`.claude/agents/loop-verifier.md`	the read-only reviewer — fails any patch that scope-creeps
`opencode.json.example` · `skills/loop-task.md`	the OpenCode twin — same skill text, event heartbeat

Loop Ready checklist

Provable success condition — the source's own spec: reproduces before the fix, doesn't after, regressions pass
Run limit — 20 runs, 2 reproduction attempts/ticket (the attempt cap is the source's own rule)
Spine written first, committed empty
Report-only (L1) start — drafts the fix in a throwaway worktree, never opens a PR
Human gate placed — the source's own review-ready checkpoint
One log line per beat, no silent runs
Kill switch — loop-pause-all in loop-budget.md

How this relates to the course

This kit uses Forward Future's Loop #16 almost unmodified. Its own rule — "do not fold unrelated refactors into the patch" — matches this repo's own loop-constraints.md rule ("never refactor unrelated code — one fix per run") almost word for word, and the loop-verifier enforces it as an automatic FAIL. Its "two serious attempts, then say so" rule is the doom-loop bound from infinite-loops.md, scenario 1, already built into the original design rather than added by this library.

Full attribution: resources/sources.md.