Repository navigation
Description
Follow-up to #4611 (PR #5470 leaves it out of scope).
Problem
On the terminal path, /planning:plan tells the session to hand-write the reconfirm resolution of a superseded-by-plan row as reconfirmed at plan approval: <new>; was: <old>:
plugins/planning/skills/plan/SKILL.md:275plugins/planning/skills/interview/context/loop.md:238
Before PR #5470, export-ledger wrote that text for a page accept on a superseded row; the named-field grammar replaces it, so only the hand-written path still produces it. import-ledger has no reader for it and treats the whole string as the row's own text, so the new and old answers do not come back as a structured pair. Pinned by the superseded-accept fixture in plugins/planning/surface/tests/fixtures/ledger-legacy/: the imported terminal decision is own with text equal to the full string, commits empty, and no was.
Scenario (c) of #4611 (a superseded row reconfirmed by an accept keeps its was answer as structure) therefore round-trips on the page path once #5470 merges and is still lost on the hand-written terminal path.
Options for triage
- Change both skill texts to hand-write the named-field form the exporter writes for the same state, so import restores it.
- Teach the importer to read the hand-written
reconfirmed at plan approval: <new>; was: <old>form into the same structured state as option 1. - Both, with the importer reader kept as a legacy grammar.
The importer must keep accepting the current text either way, because ledgers with it already exist.
Acceptance criteria
- A ledger row a session hand-writes for a reconfirmed superseded row imports to the same question state as the page path's accept on a superseded row (decision,
was, proposal pair). - The
superseded-acceptlegacy fixture still imports unchanged, or its expected state is updated with the reason stated in the PR. -
plan/SKILL.mdandinterview/context/loop.mddescribe one form;interview-defenses.test.shdigest re-pinned ifloop.mdchanges. - A test round-trips the hand-written form.
Out of scope
The page export grammar itself (#4611) and the two held-row losses named in PR #5470.
Activity
fix(planning): one escaped named-field ledger row grammar (#5470)
test(planning): re-pin the loop.md open-question register digest
kyle-sexton commentedon Oct 1, 2026
Owner decision (2026-10-01): approved the recommendation ("go with recommended", the owner's reply to the batch decision page that listed this question and its recommendation).
Question: how should #5534 be finished? Chosen option A: an agent finishes PR #5627 (change both skill texts to hand-write the named-field row the exporter writes), with the PR brought current with main, the failing test fixed,
Closes #5534in the body,ci-statusgreen, then merged through the normal gate. No legacy-form importer reader (option C is not adopted; file it as a follow-up only if wanted).Next: the issue is labeled
agent-ready; an agent takes PR #5627 to a greenci-statusand merges it.