feat(ce-plan): add U-IDs and origin trace to plan template

[tmchow]

Summary

Plans now surface origin Actors / Key Flows / Acceptance Examples in Requirements Trace, carry forward the Deep-product Scope Boundaries split when origin supplies it, and assign stable plan-local U-IDs to Implementation Units that survive deepening reorders. Test scenarios can link to origin Acceptance Examples (Covers AE1.), and ce-work / ce-work-beta recognize U-IDs in blockers, verification, and task labels. End-to-end traceability now runs from a ce-brainstorm requirements doc through ce-plan's output and into ce-work's execution surface.

This is the second half of the work PR #629 opened. PR #629 added the upstream IDs (R/A/F/AE) and the Deep-product brainstorm tier. This PR completes the plan-side artifact and the execution-side recognition.

What changes

ce-plan template (Phase 4.2). Requirements Trace gains optional **Origin actors:** / **Origin flows:** / **Origin acceptance examples:** sub-blocks, included only when the upstream brainstorm supplies them. Scope Boundaries supports a conditional three-way split for Deep-product origins:

• Deferred for later (carried — product/version sequencing)
• Outside this product's identity (carried — positioning rejection)
• Deferred to Follow-Up Work (plan-local — implementation work split across other PRs/issues/repos)

The plan-local subsection is renamed from "Deferred to Separate Tasks." "Task" overlapped with TaskCreate / TaskList tooling, and the section's contents are PRs, issues, or repos rather than tasks.

Stable U-IDs for Implementation Units. Each unit's heading is - [ ] U1. **[Name]**, paralleling the R / A / F / AE format from PR #629. The stability rule is explicit and restated at three locations because deepening (Phase 5.3) is the most likely accidental-renumber vector: U-IDs are never renumbered, gaps are fine, new units take the next unused number. Reordering preserves IDs in their new order.

AE-link convention for test scenarios. When a test scenario directly enforces an origin Acceptance Example, prefix it with Covers AE1. (or Covers F2 / AE3.). Sparse by design — most test scenarios are finer-grained than AEs and don't link. The link is the mechanism by which AE's disambiguation propagates into enforcement, symmetric to brainstorm's Covers: R-IDs upward trace.

Planning-side Alternatives rule. Alternatives differ on how the work is built: architecture, sequencing, integration pattern, rollout strategy. Tiny implementation variants belong in Key Technical Decisions. Product-shape alternatives go back to ce-brainstorm rather than re-litigating product questions during planning.

Phase 5.1 finalization checklist and deepening-workflow.md checklists. Both gain origin-trace preservation checks using judgment-call phrasing ("origin R / F / AE that affects implementation"), not mechanical "every ID appears." All origin-related checks are guarded by "if origin exists" so the no-origin path skips them naturally. deepening-workflow.md adds an explicit U-ID stability warning to the Implementation Units checklist and to Step 5.3.7's Allowed and Do-not lists. Deepening must never renumber existing U-IDs even when reordering or "tidying."

ce-work and ce-work-beta. Both recognize U-IDs alongside R / A / F / AE in blockers, deferred-work notes, task summaries, and final verification. When creating tasks from plan units, the U-ID becomes the task subject prefix (e.g., "U3: Add parser coverage") so blockers like "U3 blocked on missing fixture" anchor unambiguously to the source unit. Synced atomically per the repo's stable/beta convention.

Design notes

• Conditionality is first-class. Every origin-derived structure has a stated trigger, a stated no-origin fallback, and HTML-comment guidance so the template doesn't render broken sections in the no-origin path. ce-plan must remain useful when no upstream brainstorm exists — piecemeal use is by design.
• No new tier or namespace beyond U. No "Deep-product-plan" classification; planning answers HOW. No Q-IDs for open questions. No rename of "Implementation Units." The brainstorm-side decisions about avoiding ID-namespace sprawl apply equally here.
• The three-way Scope Boundaries split triggers on observable origin content (presence of an Outside this product's identity subsection in origin), not on a propagated tier flag. The plan never has to know the origin's tier classification — it just inspects what the origin doc actually contains.
• The AE-link convention says "should when directly enforces," not "may." "May" is too weak; agents skip optional rules under pressure. The "directly enforces" qualifier gates the rule on a real condition while still mandating compliance when it holds.
• Three distinct concepts in Scope Boundaries, not two collapsed. Origin's Deferred for later (product sequencing), origin's Outside this product's identity (positioning), and the plan-local Deferred to Follow-Up Work (implementation split) all mean different things. An earlier draft conflated the first and third; reviewing the brainstorm-side intent restored the distinction.
• ce-work and ce-work-beta synced. Sync decision stated in the U5 commit body: "Propagated to beta — shared traceability contract." Nothing about the experimental delegate mode interacts with ID semantics.

Validation

bun test — 831 pass, 0 fail. bun run release:validate — clean (50 agents, 45 skills, no drift).

The plan doc that drove this work is included in the diff: docs/plans/2026-04-21-001-feat-ce-plan-traceability-loop-plan.md. It's also the first plan written using the new conditionality rule (no origin doc, so no Origin Trace sub-blocks, scope boundaries collapses to single list). Demonstrates the no-origin fallback works.

---