# Part 6 — The Common Harness: Capstone Bundle

The synthesis of the whole series. This bundle gives you the **complete platform
skeleton** plus the four operating documents that turn it into a rollout.

> QRefAI *AI Coding Field Guide*, Part 6. Everything in Parts 2–5 was a component;
> this is where they assemble into one repository you can actually stand up.

```
.
├── dev-platform/                       # 1. ★ THE capstone — complete repo skeleton
│   ├── scripts/compile.sh              #    author-once → compile to both targets (RUNS; verified)
│   ├── instructions/ skills/ agents/   #    vendor-neutral SOURCE folders (empty-but-valid)
│   ├── mcp-servers/ policies/ workflows/
│   ├── .claude-plugin/ plugins/        #    Claude distribution
│   ├── templates/repo-template/        #    what a new product repo inherits
│   ├── .github/workflows/compile-check.yml  # CI drift gate
│   └── README.md                       #    the architecture map + quick start
├── claude-copilot-mapping.xlsx         # 2. Full capability mapping (+ .csv)
├── claude-copilot-mapping.csv
├── phased-asset-checklist.md           # 3. Phases 0–6 build checklist (Notion-importable)
├── ten-things-quickstart.pdf           # 4. One-page printable sprint board
└── rollout-scorecard.xlsx              # 5. Phases A–D gate-criteria scorecard
```

## The five pieces

1. **`dev-platform/` skeleton** — the single most valuable download in the series.
   Every folder from the Q6.1 tree, every file empty-but-valid, and a
   `scripts/compile.sh` that **actually runs**: it compiles the vendor-neutral
   source folders into a Claude target (`.claude-plugin/`) and a Copilot target
   (`.github/`), with a `--check` mode for CI drift-gating. Fill the `<PLACEHOLDER>`s
   from the Parts 2–5 templates and you have a working platform repo.

2. **Claude↔Copilot mapping sheet** — the precise 1:1 correspondence the compiler
   implements, as a working file. The "Authored once as" / "Compiles to" columns are
   the contract between your source folders and the two targets.

3. **Phased asset checklist** — Phases 0–6 (Foundations → Operate) with interactive
   checkboxes and an exit criterion per phase. This governs *what you build*.

4. **Ten-things quick-start** — the highest-leverage subset as a one-page printable
   sprint board, colour-coded by harness layer. Pin it up; work top to bottom.

5. **Rollout scorecard** — Phases A–D (Build → Pilot → Department → Org-wide) with
   *measurable* gates (<5% blocked sessions, ≥99.5% uptime, >25% self-reported gain,
   …) and a Status dropdown. This governs *how you deploy* what you build.

## How to use the bundle

```bash
# 1. Stand up the skeleton as your platform repo.
cd dev-platform && scripts/compile.sh         # proves it compiles on the empty tree
# 2. Work the phased-asset-checklist.md to fill in real content (Parts 2–5 templates).
# 3. Pin ten-things-quickstart.pdf for the team; do those ten first.
# 4. Deploy using rollout-scorecard.xlsx — advance a phase only when its gates are green.
```

## The whole series in one sentence

A coding agent's quality is set less by the model than by the **harness** around it
— context, capability, control, governance — and this bundle is that harness as a
buildable, distributable, governable repository.

## Notes

- **`compile.sh` is a verified stub:** the folder contracts, ordering, and `--check`
  drift gate are real and run; the per-file frontmatter transforms are marked `TODO`.
- **The generated `build/` is intentionally absent** — it's an output of `compile.sh`,
  not a source artifact. Run the script to produce it.
- **Spreadsheets** ship with a Status/RAG dropdown and zero formula errors.
- Consistent with Parts 2–5: `//`-key JSON docs, real-directory layout, README index.

_Snapshot of the mid-2026 consensus. Treat the harness as a living product, and
re-verify vendor specifics before relying on them — this field moves monthly._