Slate v2 Deletion Closure Protocol Hardening Consensus Plan

RALPLAN-DR Summary

Principles

1. closure claims must match the exact audited glob, not a vibe-equivalent package label
2. parent buckets stay open until every child bucket is closed or explicitly skipped
3. roadmap truth should be refreshed from ledger truth, not improvised from recent memory
4. process rules belong in one durable protocol owner, not scattered across random closeout notes

Decision Drivers

1. the recent packages/slate-react/test/** batch was closed correctly, but the control docs omitted the still-open packages/slate-react/src/** deletion pressure from the remaining-open summaries
2. the current roadmap stack already has clean owner boundaries that should be preserved
3. future deletion batches need a mechanical gate, not another “be more careful” speech

Viable Options

Option A: Encode the new rules directly into master-roadmap.md and release-file-review-ledger.md

Pros:

• no new top-level protocol doc
• fewer files touched

Cons:

• clutters control docs with process mechanics
• makes the rules harder to find and easier to drift

Option B: Add one dedicated deletion-closure protocol doc, then wire roadmap/ledger/command docs to obey it

Pros:

• one durable process owner
• keeps roadmap/ledger docs tighter
• easiest place to centralize gates, examples, and wording bans

Cons:

• adds one more reference doc
• requires a small amount of cross-doc wiring

Option C: Only tighten reconsolidate-roadmap.md

Pros:

• smallest immediate patch
• minimal doc churn

Cons:

• too weak
• does not protect non-command-driven edits
• still leaves package-vs-glob closure semantics implicit

Option D: Only tighten refresh-file-review-ledger.md

Pros:

• hits the file-truth owner directly
• smaller than a broader protocol rollout

Cons:

• still leaves roadmap/front-door wording vulnerable
• too easy for operator edits to bypass

Chosen Option

• Option B

Why chosen:

• it adds one real process owner instead of stuffing protocol sludge into docs that already have tighter jobs

Task Statement

Harden the Slate v2 roadmap/process stack so deletion-closure claims must follow the exact audited glob hierarchy, with parent/child promotion gates, residual-open disclosure, frozen-inventory reconciliation, and existing proof-surface / lane-id citations where behavior lanes demand them.

Scope

In Scope

• a new deletion-closure protocol doc under docs/slate-v2/
• roadmap wiring in:
  • master-roadmap.md
  • overview.md
  • release-file-review-ledger.md
  • commands/reconsolidate-roadmap.md
  • commands/refresh-file-review-ledger.md
• explicit parent/child closure-tree semantics
• wording rules that ban package-wide closure claims when only a child glob is closed
• immediate correction of the current Slate React overclaim as the first proving case
• protocol treatment for historical closeouts that predate the new required closeout shape

Out of Scope

• executing the full packages/slate-react/src/** deleted-source closure
• reopening the finished packages/slate-react/test/** family batch
• changing verdict docs
• redesigning the entire roadmap stack beyond deletion-closure protocol needs

Decision

Add one new protocol owner:

• /Users/zbeyens/git/plate-2/docs/slate-v2/references/deletion-closure-protocol.md

This document owns the mechanics:

1. batch naming must equal the exact audited scope
2. package-level closure requires parent/child reconciliation
3. closeout notes must disclose residual-open sibling buckets
4. refreshes must reconcile against a frozen deleted inventory, not an ad hoc rerun
5. behavior-sensitive deletion batches must cite existing proof surfaces and existing lane ids only
6. umbrella wording is banned unless the parent bucket is actually reconciled

Then wire the control stack to obey it:

• master-roadmap.md
  • references the protocol in Tranche 4 and Batch Exit Rule
• release-file-review-ledger.md
  • carries mandatory package-level parent rows for split deletion packages
  • records that child-bucket closure does not auto-promote the parent
• overview.md
  • uses protocol-safe wording in Current Read
• commands/reconsolidate-roadmap.md
  • adds the frozen-inventory gate before control-doc refresh
• commands/refresh-file-review-ledger.md
  • enforces the same frozen-inventory and promotion-gate rules on the ledger owner path

The first consumer of the protocol is the current Slate React drift:

• packages/slate-react/test/** stays closed
• packages/slate-react/** stays open until src/** residue is classified

Proposed Protocol

Rule 1: Batch Name Equals Audited Scope

Allowed:

• Slate React deleted test-family closure: packages/slate-react/test/**
• Slate React deleted source-family closure: packages/slate-react/src/**

Not allowed:

• Slate React closure
• Slate React deletion closure

Meaning:

• the plan title, closeout title, and ledger wording must preserve the exact audited scope

Rule 2: Parent / Child Closure Tree

Every package with split deletion buckets must define a closure tree.

The seeded tree must come from the frozen deleted inventory, not memory.

Current Slate React example from the live deleted inventory:

• packages/slate-react/**
• packages/slate-react/test/**
• packages/slate-react/src/hooks/android-input-manager/**
• packages/slate-react/src/chunking/**
• packages/slate-react/src/components/restore-dom/**
• remaining deleted packages/slate-react/src/hooks/* files derived from the frozen inventory
• remaining deleted packages/slate-react/src/components/* files derived from the frozen inventory
• single-file residue:
  • packages/slate-react/CHANGELOG.md
  • packages/slate-react/src/utils/environment.ts
  • packages/slate-react/src/custom-types.ts
  • packages/slate-react/src/@types/direction.d.ts

The grouped “remaining deleted ... files derived from the frozen inventory” rows above mean: derive the exact leftover deleted paths in those subtrees from the frozen inventory and enumerate them in the parent/child matrix before any closure wording changes.

Parent promotion rule:

• parent stays open until all children are:
  • closed
  • or explicit skip

Rule 3: Promotion Gate

Closing a child bucket does not close the parent bucket.

Example:

• allowed:
  • packages/slate-react/test/** is closed
  • Slate React test-family is closed
• forbidden unless the parent reconciles:
  • Slate React is closed
  • removing slate-react from the remaining-open roadmap bullets

Rule 4: Residual-Open Disclosure

Every deletion-family closeout note must end with:

• Scope closed
• Sibling buckets still open
• What this batch does NOT close

If there are no open siblings, the note must say that explicitly.

Historic closeout handling:

• pre-protocol closeouts are grandfathered
• if a historic closeout is actively refreshed during the rollout, it must be upgraded to the new closeout shape
• otherwise the required residual-open section applies to new closeouts and refreshed closeouts only

Rule 5: Frozen Inventory Gate

Before editing control docs, capture the deleted-path inventory once from the target repo and freeze it for the batch.

git -C /Users/zbeyens/git/slate-v2 diff --diff-filter=D --name-only -- <package-or-scope>

Then derive the exact child inventory from that frozen set and classify each path as:

• already closed in ledger
• closed by this batch
• still open
• explicit skip

Do not rerun the raw deleted diff later and treat it as a new source of truth.

If any sibling source bucket in the frozen inventory is still open, top-level docs must keep the parent package visible in remaining-open language.

Rule 6: Existing Proof-Surface / Lane Citations Only

Deletion closure is not just file accounting.

When the restored/cut behavior touches named runtime lanes, the batch must cite:

• existing proof surfaces from
  • true-slate-rc-proof-ledger.md
  • release-file-review-ledger.md
• existing specialist lane ids from
  • docs/slate-browser/proof-lane-matrix.md

Example:

• Editable / focus / composition changes
  • proof surfaces:
    • runtime/browser
    • public surface
  • lane ids:
    • yarn test:slate-browser:ime:local
    • yarn test:slate-browser:dom

No freeform owner taxonomy like selection-sync.

Rule 7: Umbrella Wording Ban

Allowed:

• ``packages/slate-react/test/** is closed
• Slate React test-family is closed

Not allowed:

• Slate React deletion review is closed

unless the parent packages/slate-react/** row reconciles to zero open children.

Doc Ownership Plan

New Owner

• docs/slate-v2/deletion-closure-protocol.md
  • owns deletion-closure mechanics and examples

Updated Owners

• docs/slate-v2/master-roadmap.md
  • references the protocol as Tranche 4 entry/exit discipline
• docs/slate-v2/release-file-review-ledger.md
  • carries the actual parent/child status truth
• docs/slate-v2/overview.md
  • mirrors only protocol-safe front-door wording
• docs/slate-v2/commands/reconsolidate-roadmap.md
  • enforces the frozen-inventory gate and refresh checklist
• docs/slate-v2/commands/refresh-file-review-ledger.md
  • enforces the same frozen-inventory and promotion-gate rules on the file-truth path

Execution Phases

Phase 0: Patch The Current Overclaim

Goal:

• immediately correct the current omission-based Slate React overclaim in the control docs

Exit:

• packages/slate-react/test/** remains closed
• packages/slate-react/** is still visibly open until src/** is reconciled

Phase 1: Add The Protocol Owner

Goal:

• write deletion-closure-protocol.md

Must include:

• exact-scope naming rule
• parent/child closure-tree rule
• promotion gate
• residual-open section contract
• frozen inventory gate
• existing proof-surface / lane-id citation rule
• historic closeout grandfathering rule
• wording ban examples

Phase 2: Wire The Control Docs

Goal:

• make the protocol unavoidable in the normal roadmap refresh path

Required edits:

• master-roadmap.md
• release-file-review-ledger.md
• overview.md
• commands/reconsolidate-roadmap.md
• commands/refresh-file-review-ledger.md

Phase 3: Seed The First Package-Level Closure Tree

Goal:

• add the first explicit parent/child matrix for packages/slate-react/**

This matrix must be exhaustive against the frozen deleted inventory, including:

• packages/slate-react/test/**
• packages/slate-react/src/chunking/**
• packages/slate-react/src/components/restore-dom/**
• packages/slate-react/src/hooks/android-input-manager/**
• remaining deleted hooks/components grouped only from the frozen inventory
• package-root and single-file residue

This proves the protocol on a real package and removes ambiguity about what is still open.

Acceptance Criteria

1. one new protocol doc exists and owns deletion-closure mechanics
2. master-roadmap.md, overview.md, release-file-review-ledger.md, and both command docs all reference or enforce the protocol in their own role-appropriate way
3. package-wide closure language is impossible to justify without a reconciled parent row
4. closeout notes have a required residual-open section for new and refreshed closeouts, while historic closeouts are explicitly grandfathered unless touched
5. the refresh path now requires a frozen deleted-inventory classification step before top-level wording changes
6. the current Slate React overclaim is corrected as part of the rollout
7. specialist verification references use existing proof surfaces and existing lane ids only
8. the seeded packages/slate-react/** tree is exhaustive against the frozen deleted inventory, not a representative sample

Verification

Planning artifact verification:

• bunx prettier --check docs/plans/2026-04-09-slate-v2-deletion-closure-protocol-hardening-consensus-plan.md

Execution-phase verification should include:

• bunx prettier --check on the touched roadmap/process docs
• rg -n "packages/slate-react/test/\\*\\*|packages/slate-react/src/\\*\\*|deletion-closure-protocol|remaining open deletion|frozen inventory" docs/slate-v2/master-roadmap.md docs/slate-v2/overview.md docs/slate-v2/release-file-review-ledger.md docs/slate-v2/commands/reconsolidate-roadmap.md docs/slate-v2/commands/refresh-file-review-ledger.md docs/slate-v2/deletion-closure-protocol.md

If the rollout also corrects live Slate React package wording:

• the frozen deleted inventory artifact must live in /Users/zbeyens/git/plate-2/docs/plans/2026-04-09-slate-v2-deletion-closure-protocol-rollout.md under a Frozen Deleted Inventory section
• the seeded packages/slate-react/** parent/child matrix must have zero unmatched paths against that frozen inventory
• the control docs must still keep the parent package visibly open until every child row is closed or explicit skip

Risks And Mitigations

• owner-boundary duplication
  • mitigation: the protocol doc owns schema/invariants only; live status stays in release-file-review-ledger.md, lane ids stay in docs/slate-browser/proof-lane-matrix.md
• stale diff reruns changing truth mid-refresh
  • mitigation: freeze the deleted inventory once per batch and reconcile from that artifact only
• historic closeouts instantly failing the new contract
  • mitigation: grandfather untouched historic closeouts and require the new closeout shape only for new or refreshed closeouts
• under-enumerated child buckets leaking package-wide closure again
  • mitigation: require an exhaustive seeded parent/child matrix against the frozen inventory before any parent-level wording change

ADR

Decision

Create one dedicated deletion-closure protocol doc and wire the control stack to obey it.

Drivers

• exact audited scope matters more than package-name convenience
• parent/child promotion needs a hard gate
• refresh commands need a frozen inventory-reconciliation step

Alternatives Considered

• embed all rules directly in roadmap + ledger
• only tighten the reconsolidation command
• only tighten the ledger refresh command

Why Chosen

• best balance between clarity, durability, and owner-boundary discipline

Consequences

• one extra reference doc exists
• both refresh commands now share one protocol instead of drifting
• control docs stay cleaner because mechanics move out of them
• future batch overclaims should become visibly invalid instead of easy to type

Follow-Ups

• execute this protocol rollout before the next packages/slate-history/test/** deletion batch
• use packages/slate-react/** as the first seeded package-level closure tree

Agent Roster And Staffing

Available useful agent types for follow-up execution:

• executor
• architect
• critic
• verifier
• writer

Ralph Path

Best follow-up is $ralph, not $team.

Why:

• this is a small, tightly coupled docs/process patch
• parallel lanes would just create merge noise in the same control docs

Suggested Ralph work:

1. add deletion-closure-protocol.md
2. patch roadmap/ledger/overview/command docs
3. correct the current Slate React overclaim
4. rerun doc verification

Team Path

Only use $team if you insist on splitting:

• lane 1: protocol doc draft
• lane 2: roadmap/overview wording patch
• lane 3: ledger/command patch

This is probably slower than Ralph because the write surfaces overlap.