# Cross Layer Kickoff Review

> How to gate a multi-agent kickoff batch before calling it a real vertical slice

- Skill: `majiayu000/cross-layer-kickoff-review` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds add majiayu000/cross-layer-kickoff-review`
- Raw SKILL.md: https://api.skillmd.com/api/skills/majiayu000/cross-layer-kickoff-review/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: majiayu000 (https://skillmd.com/u/majiayu000)
- Updated: 2026-09-09
- Page: https://skillmd.com/skills/majiayu000/cross-layer-kickoff-review

---


## Context

Use this when host, services, and web UI are being built in parallel and the team wants an honest readiness verdict instead of placeholder optimism.

## Pattern

1. Verify the toolchain first: if the solution does not build or tests cannot run, the batch is not review-complete.
2. Compare the host and Web UI message vocabulary directly: request names, response names, capability labels, and seeded constants must agree exactly.
3. Separate placeholder seams from integrated seams. A scaffold page or stub handler is useful, but it does not satisfy a vertical-slice success criterion.
4. Check the first failure contract, not just the happy path. Empty input, unsupported messages, and unplaced-reason codes should already be stable enough to assert against.
5. Trace each claimed success criterion to one real code path, not just prose or helper tests.

## Good signs

- One real request can leave the UI, cross the bridge, hit application logic, and return shaped data without a placeholder response.
- Seeded material/settings values match across host, services, and web contracts.
- Review evidence includes at least one runnable verification command, not just static inspection.

## Anti-patterns

- Declaring a vertical slice complete while file-open/import/nesting still return `not-ready`
- Letting host and UI invent different bridge names during parallel work
- Treating spec-helper tests as proof that the implemented seam already works

