Editor-Behavior Replan Next Batch

Goal

Refresh the remaining editor-behavior backlog after release-prep was removed from roadmap truth, and name the next executable lane without falsifying the formal backlog order.

Decision

• Keep the formal backlog exactly aligned with docs/editor-behavior/markdown-parity-matrix.md.
• Treat search / find-replace as the next executable lane.
• Search is now also the formal top backlog item because the docs-only polish slice was completed.
• Search belongs to the broader editor-behavior backlog as a cross-surface lane, not to the markdown-native lane specifically.
• Do not launch Ralph from the old major PRD/test-spec.

Truth Backlog

This remains the source-of-truth order:

1. editor-behavior-wide search / find-replace product lane
2. toggle rewrite lane
3. code-drawing / Excalidraw lane
4. collaboration / editor-only lane
5. deferred feature-gap follow-up:
   • richer date MDX payloads beyond the current plain <date>value</date> contract
   • richer media/embed source-entry or provider-metadata law beyond the current url+attribute contract
6. streaming improvements only when a real lane needs them

Next Executable Lane

search / find-replace

Why:

• its behavior law is already locked in docs/editor-behavior/markdown-editing-spec.md
• its deferred rows already exist in docs/editor-behavior/editor-protocol-matrix.md
• it is a better next Ralph batch than spending the next batch on docs polish
• it is better grounded than jumping straight into the toggle rewrite lane

Non-Decision

The old split is gone. After the docs-only polish pass, search / find-replace is both the formal top backlog item and the best next executable lane. That does not make it a markdown-native lane; it is a cross-surface editor-behavior lane.

Search Scope For The Next Lane

The next search planning artifacts must cover the locked search law, not just the current highlight demo:

• current-file search
• seeded search from selection
• next / previous match
• jump-to-selection
• replace
• shared navigation feedback for search jumps
• outline header search

Current implementation reality is still thin:

• packages/find-replace/src/lib/FindReplacePlugin.ts is highlight plumbing
• apps/www/src/registry/examples/find-replace-demo.tsx is a demo toolbar over that plumbing
• packages/find-replace/src/lib/decorateFindReplace.spec.ts covers decoration, not a full search surface

Required Next Artifacts

Before any Ralph launch for search, create:

• .omx/plans/prd-editor-behavior-search.md
• .omx/plans/test-spec-editor-behavior-search.md

Those artifacts must map directly to:

• docs/editor-behavior/markdown-editing-spec.md search law
• docs/editor-behavior/editor-protocol-matrix.md deferred search rows

Required Verification For The Search Lane

The search mini test-spec must name concrete verification lanes for:

• package tests
• app integration tests
• browser verification
• shared navigation-feedback coverage

The browser proof must cover at least:

• open current-file search
• seed search from selection
• move next / previous
• replace the active match without corrupting structure
• jump to selection
• search-target navigation feedback

Command-Doc Drift To Fix

These docs still carry stale post-major routing or stale next-order wording:

• docs/editor-behavior/commands/README.md
• docs/editor-behavior/commands/replan-next-batch.md
• docs/editor-behavior/commands/launch-next-ralph-batch.md

What changes:

• replan-next-batch.md should stay generic as the replan operator flow
• launch-next-ralph-batch.md should hand off search execution to the new search PRD/test-spec instead of the old major artifacts
• README.md should reflect honest post-major routing

Replan Acceptance Criteria

This replan is only valid if:

• the truth backlog matches the parity matrix exactly
• search is named as the next executable lane without reordering the formal backlog
• the search lane is explicitly scoped to locked law plus deferred protocol rows
• the next search handoff requires new search-specific planning artifacts
• the command docs are recognized as stale where they still encode old major-era routing

Pre-Mortem

1. Search scope drifts into cross-file or app-shell behavior too early. Mitigation: keep the next search PRD limited to locked document-level law.
2. Outline header search gets awkwardly split from shared search/navigation feedback. Mitigation: keep both in the same search-lane planning pass unless there is a written reason to separate them.
3. Search invents one-off highlight/flash behavior instead of reusing the shared navigation-feedback primitive. Mitigation: require explicit navigation-feedback coverage in the mini test-spec and browser proof.