docs(decisions): add COR-040 process open region

[aleskalfas]

Closes #417.

What

Ships COR-040 — "A process may declare an open region", the last un-shipped slot in the process-substrate parking lot (EPIC #237). An open region is a single state with no internal edges + applies_to-scoped invariants + one COR-033-P4 exit gate. Boundary-enforcement is delivered as exit-gate predicate composition (the engine never blocks a move because of an invariant), so COR-035's report-only default is preserved verbatim; mutation-time fail-closed stays the binding's command-wrapper job; liveness is the binding's gate-design problem (status surfaces non-progress).

Additive un-defer of COR-033-P6 + COR-035's deferred applies_to / boundary-enforcement; supersedes nothing. Includes the minor version bump 1.132.0 → 1.133.0 for the new accepted-COR principle (PRJ-002).

Doc impact

None in this PR. The .pkit/process/README.md deferred→core row flip (open-region + applies_to / boundary-enforcement) lands with the engine task #418, not here — this PR adds the decision record and the version bump only. No migration owed (pure addition; pkit migrations check-diff clean).

Test plan

N/A for a decision-authoring PR. Review coverage: architect + critic (design) and methodology-reviewer (the authored record).

Reviews

• Design reviewed by architect + critic (2026-07-01); the decisive fix (open region = single state; enforcement = gate composition, not move-blocking invariants) is baked in.
• COR reviewed by methodology-reviewer — PASS, 0 blockers; additive-un-defer framing confirmed, report-only preserved verbatim.

Scope

This is Task #417 (author + accept the COR). The engine + shape contract + engine-strategy ADR is #418 (unblocked by this); grounding on trip-planning build is #419.

🤖 Generated with Claude Code

[aleskalfas]

Reviewer agent (local, reviewer): CHANGES_REQUESTED

Reviewed PR #423 (docs(decisions): add COR-040 process open region, branch docs/417-author-the-open-region-cor, closes #417) against the project-management capability's conventions. One gate-blocking finding forces the verdict.

Hard-reject (gate-blocking)

• PR body is missing a ## Doc impact section. The pr-body convention in git-conventions.yaml marks "PR body contains a ## Doc impact section" as [validation-severity:hard-reject]. The body carries ## What, ## Reviews, and ## Scope only — no ## Doc impact. Notably the closing issue [Docs] Author the open-region COR (additive un-defer of COR-035 applies_to + COR-033 P6) #417 does have a Doc impact note (the .pkit/process/README.md row flip is deferred to the engine task [Task] Implement the open-region slot in the shape contract + engine, and author the engine-strategy ADR #418); that belongs mirrored into the PR body per the convention. Add a ## Doc impact section (even "None — the deferred→core row flip lands with the engine task [Task] Implement the open-region slot in the shape contract + engine, and author the engine-strategy ADR #418, not here") to clear the gate. ([project-management:DEC-013-branch-and-pr-conventions], git-conventions.yaml pr-body)

Warnings (non-blocking, worth addressing)

• Conv-type divergence: docs vs decision for a decision-landing PR. git-conventions.yaml's branch-name block states that a PR landing a decision record (COR/PRJ/DEC/ADR) "takes the decision conv-type explicitly … overriding the closing Task's mapped type … its PR is decision(<scope>): …". This PR lands COR-040 but titles/branches as docs. The literal hard-reject rule (conv-type matches pr_type_mapping, docs→docs) is satisfied, so this is not gate-blocking — but there's an internal inconsistency: the sole authoring commit is decision(core): add COR-040 … while the PR title (and thus the squash-merge subject) is docs(decisions): …, so COR-040 would land on main under a docs commit. Recommend retitling to decision(core): add COR-040 — process open region (and, ideally, the branch to decision/417-…) to match the shipped content and the commit. ([project-management:DEC-013-branch-and-pr-conventions] / COR-008)
• No ## Test plan section. The empty-checkbox-skeleton rule (DEC-031) targets a present-but-empty section, so outright absence isn't a mechanical hard-reject here; for a decision-authoring PR a test plan is arguably N/A. Consider a one-line ## Test plan noting review coverage (architect/critic/methodology-reviewer) for completeness.

Passing criteria (noted, not blocking)

• Branch shape matches ^(…|docs|…)/[0-9]+-<slug>$. Closing-issue link present (Closes #417 + populated closingIssuesReferences). Issue [Docs] Author the open-region COR (additive un-defer of COR-035 applies_to + COR-033 P6) #417 classification complete (type:docs, priority:Medium, workstream:workflow).
• Surface-change discipline: pure addition (new COR + version bump 1.132.0→1.133.0 + requires_backbone ceiling broaden); no renames/removals, no schema_version bump, no subtree restructure — no migration owed, consistent with the body's claim (COR-010, PRJ-002).
• No-shared-files: the edits under .pkit/decisions/core/, .pkit/VERSION, and .pkit/adapters/claude-code/package.yaml are core-owned, but this is project-kit's own self-hosting source repo — core is authored here by design, so COR-001 is not violated.

Resolve the missing ## Doc impact section and re-request review; the conv-type warning is strongly recommended but not gate-blocking.

[aleskalfas]

Reviewer agent (local, reviewer): APPROVED

Review of PR #423 (docs(decisions): add COR-040 process open region), closing issue #417, base main.

Warnings (non-blocking):

• PR title conv-type vs. the decision-landing override. The title is docs(decisions): …, which mechanically satisfies the mapping — issue [Docs] Author the open-region COR (additive un-defer of COR-035 applies_to + COR-033 P6) #417 carries type:docs, and pr_type_mapping sends docs → docs, so the hard-reject branch/title rules in git-conventions.yaml pass. But this PR lands a decision record (COR-040), and the branch-name convention's decision note ([project-management:DEC-013-branch-and-pr-conventions], echoing [COR-008]) says such a PR "takes the decision conv-type explicitly … overriding the closing Task's mapped type." The PR's own commits already follow that (decision(core): add COR-040 …, decision(core): accept COR-040 …). Because merge is squash-with-delete-branch and the squash subject = PR title, the base-branch commit will read docs(decisions): … rather than decision(core): …, diverging from the project's decision conv-type for decision-landing changes. Worth realigning the title to decision(core): … before merge, but the schema's severity model makes this a warning, not a gate block.

Criteria that pass, worth noting given the self-host context:

• No-shared-files ([COR-001]). The diff writes .pkit/decisions/core/COR-040-…md and .pkit/adapters/claude-code/package.yaml — normally core/adapter-owned trees an adopter must not touch. This repo is the methodology source that self-hosts, so authoring a core decision record and auto-broadening the adapter's requires_backbone ceiling is this repo's sanctioned product work, not an invariant breach. No violation.
• Surface-change discipline ([COR-010]). Pure addition — a new COR file plus a .pkit/VERSION bump (1.132.0 → 1.133.0) and the paired requires_backbone broaden. No renames/removals, no schema_version bump, no subtree restructure; the record itself asserts additive/no-migration and the body reports pkit migrations check-diff clean. No migration owed.
• Branch shape & type alignment ([project-management:DEC-013]). docs/417-author-the-open-region-cor matches ^(feat|fix|docs|test|refactor|chore|ci|decision)/[0-9]+-…$, and docs matches issue type:docs per pr_type_mapping.
• Classification complete ([project-management:DEC-012]). Issue [Docs] Author the open-region COR (additive un-defer of COR-035 applies_to + COR-033 P6) #417 carries type:docs, priority:Medium, and workstream:workflow.
• Closing-issue link ([project-management:DEC-013]). Body opens Closes #417. and closingIssuesReferences is populated; required ## Doc impact and ## Test plan sections are present with authored (non-skeleton) content.

No hard-reject or unaudited bypassable-with-audit findings → verdict follows mechanically to APPROVED.