Team Split Assessment

Plan Under Review

File: project-blueprint.md
Date: 2026-03-19

Complexity Analysis

Factor	Value	Threshold	Assessment
Total tasks	~65	15	Over — 10 runs with 6-8 tasks each, Run 5 alone has ~27 tasks across 5 sub-runs
Distinct phases with sequential dependencies	10	3	Over — Runs 0-9, fully sequential dependency chain
Distinct persona types needed	3	5	Under — front-end-engineer (primary), quality-reviewer, technical-writer
Estimated working directories	1	2	Under — single branch in ux-prototype, worktree is read-only reference
Cross-cutting concerns (handoffs, gates)	12+	2	Over — every run has entry/exit gates, Run 1 has 6 intermediate gates
Tasks with external validation (builds, tests)	15+	6	Over — lint, tsc, test, build-storybook, VRT, Playwright MCP checks per run

Qualitative factors:

Context window pressure: High. Run 1 (6 tasks × multiple files each), Run 3 (4 organisms × import rewrites), and Run 5 (5 sub-runs × new code) will each generate substantial diffs and test output.
Error blast radius: Each run modifies shared infrastructure (imports, styles, grid code). Failures must be isolated per run to avoid cascading regressions.
Natural phase boundaries: Clear checkpoint moments between runs — each run commits, validates, and hands off a stable state.

Recommendation

Split into 10 runs — the blueprint already defines the natural boundaries. Each run has distinct scope, verification gates, and deliverables. The sequential dependency chain means runs cannot be parallelized, but each run can use an internal team of 1-3 agents.

Proposed Runs

Run	Directory	Description	Tasks	Personas	Scope
0	run-0-branch-setup/	Create target branch, verify baseline	2	front-end-engineer	Trivial
1	run-1-foundation/	Styles, primitives, atoms, hook, cn rewrite	6 + 6 intermediate gates	front-end-engineer	Medium
2	run-2-visual-foundation/	DataGrid theme migration (legacy → themeQuartz)	3	front-end-engineer	Small (high visual risk)
3	run-3-organism-port/	Move sidebar, app-header, item-details, item-grid + molecules	3 groups + import rewrites	front-end-engineer	Medium-Large
4	run-4-cell-editor-upgrade/	SelectCellEditor replaces EnumCellEditor	8	front-end-engineer	Small-Medium
5	run-5-entity-grid-evolution/	Promote 7 capabilities into entity-data-grid	5 sub-runs, ~27 tasks	front-end-engineer, quality-reviewer	Large
6	run-6-item-grid-integration/	Refactor item-grid to use entity-data-grid	6	front-end-engineer	Medium-Large
7	run-7-package-api/	Barrel exports, dependencies, cleanup	3 groups	front-end-engineer	Small
8	run-8-acceptance/	Composition stories, VRT, MDX docs	3 groups	front-end-engineer, technical-writer	Medium
9	run-9-canary-refactor/	Adapt canary-refactor stories	4	front-end-engineer	Medium

Artifact Flow

Run 0 → Branch exists on jmpicnic/component-consolidation
  │
Run 1 → canary/primitives/, canary/atoms/, types/canary/hooks/, updated styles
  │       All imports canonical, existing tests pass
  │
Run 2 → DataGrid molecule with themeQuartz, VRT baselines regenerated
  │
Run 3 → All 4 callil organisms + molecules ported, imports rewritten
  │
Run 4 → atoms/grid/select/ replaces atoms/grid/enum/
  │
Run 5 → entity-data-grid with auto-publish, actions, search, pagination, toolbar, auto-height, drag-to-scroll
  │       10+ new stories, unit tests, play functions
  │
Run 6 → item-grid refactored to use entity-data-grid as base
  │
Run 7 → canary.ts barrel updated, package.json updated, deprecated files deleted
  │       Library build succeeds
  │
Run 8 → 6+ composition stories, VRT baselines, MDX docs for all components
  │
Run 9 → canary-refactor stories adapted to consolidated components

Risk Analysis

Risk	Mitigation via Splitting
Theme migration diffs every grid story	Isolated in Run 2 — if VRT reveals problems, only Run 2 needs rework
Auto-publish is complex new code	Isolated in Run 5a — failure doesn’t affect organisms ported in Run 3
Import rewrites touch 38+ files	Isolated in Run 1 with intermediate gates — failures caught per task
SelectCellEditor may break existing stories	Isolated in Run 4 — existing enum stories serve as regression gate
Item-grid refactoring may need wrapper	Run 5 proves capabilities first; Run 6 can adapt based on what Run 5 delivered
Barrel export changes affect published API	Deferred to Run 7 — all components stabilized before touching the public API

Coordination overhead: Minimal — all runs are sequential, single-persona (front-end-engineer) for most runs, filesystem-only handoffs. No cross-team coordination needed.

Open Questions and Decisions

#	Question	Options	Recommendation	Decision
1	Should Run 5 sub-runs be separate run directories or grouped?	(a) 5 separate dirs, (b) single dir with phases	(b) single dir — sub-runs share the same codebase area and persona	Grouped
2	Should Run 8 and Run 9 be combined?	(a) separate, (b) combined	(a) separate — Run 8 is acceptance of the consolidation, Run 9 is a follow-on that may surface regressions requiring iteration	Separate