write-goal

name: write-goal description: Phase 2 of pair-goal. Turns .pairgoal/UNDERSTANDING.md into a /goal-ready completion condition — one measurable end state, a stated transcript-demonstrable check, the constraints that must hold, and a turn cap — written to .pairgoal/GOAL.md plus the literal /goal line for the user to run. Use after train-orchestrator, or standalone to craft a good /goal condition from a clear improvement target.

write-goal

Phase 2 of pair-goal. Converts .pairgoal/UNDERSTANDING.md into a completion condition you can hand to Claude Code's /goal command, written to .pairgoal/GOAL.md.

What makes a good /goal condition

/goal's evaluator is a fast model that, after every turn, reads only the conversation (it runs no tools) and decides yes/no. So a strong condition has all four parts:

1. One measurable end state — a test result, a benchmark threshold, a file count, an empty queue, an exit code. Not "better" or "cleaner".
2. A stated check — how it's proven, phrased so the agent's own output demonstrates it: pytest tests/x -q exits 0, bench/run.py prints p95 <= 40ms, git status is clean.
3. Constraints that must hold — what must not change on the way there (public signatures, deps, other tests).
4. A turn clause — or stop after N rounds, so a goal that can't be met still terminates. Use ROUNDS from STATE.md (default 5).

Keep it ≤ 4000 characters. Pull the end state and signal straight from UNDERSTANDING.md's "What better means" and "Success signal" — if that signal wasn't transcript-demonstrable, fix it now or flag it loudly (the goal will otherwise loop forever, since the evaluator can never see the proof).

Process

1. Read .pairgoal/UNDERSTANDING.md. If the success signal is missing or unmeasurable, go back to the user (or train-orchestrator) — don't paper over it.
2. Draft the condition with the four parts above, on one logical line.
3. Write .pairgoal/GOAL.md (template in ../pair-goal/reference/file-formats.md): the condition, the literal /goal … line, and a note on why the check is transcript-demonstrable.
4. Set GOAL: (one-line restatement) in STATE.md, advance PHASE: goal.

The gate — the user runs /goal

You cannot execute a slash command yourself. So present the line and hand off:

Goal condition ready. Review it, then run this yourself to start the loop:

/goal <condition>

Once it's active, each turn I'll run one pair-iterate race round until the condition holds, then summarize.

Set STATUS: WAITING-USER: approve-goal. Wait for the user to approve (and possibly edit) and to actually run /goal. Only then does pair-iterate begin.

Don't

• Don't make the condition depend on something the agent won't print. If proving it requires reading a file or running a tool, make the agent run that and surface the output part of the loop's job — and say so in GOAL.md.
• Don't omit the turn clause. A goal with no cap can run away.
• Don't widen scope beyond UNDERSTANDING.md. The goal is the contract; keep it tight.