Wire the harness for GitHub Copilot

[chumleesockson]

Makes the same harness drive GitHub Copilot, since the demo will be run there. The harness was built for Claude Code (AGENTS.md + .agents/.claude skills); Copilot has its own conventions, so without this it would ignore the steering and the skills, leaving a bare model with no harness, the exact failure mode the talk warns about.

What's added (.github/ only, no conflicts with the strip-back PR)

• .github/copilot-instructions.md — auto-loaded by Copilot as repo custom instructions. Points at AGENTS.md as the steering doc and restates the essentials inline (the non-negotiable laws, the npm run verify gate, commit-per-task, the skill run order) so they're in context even if AGENTS.md isn't opened.
• .github/prompts/*.prompt.md — the four skills exposed as Copilot / prompts (/dataverse-scaffold-tables, /dataverse-create-records, /scaffold-code-app, /dataverse-add-data-source). Each is a thin pointer to the canonical .agents/.../SKILL.md, so there's no fourth copy to keep in sync, just a Copilot-shaped entry point.

Verify on the dry run (I can't confirm from here)

• That your Copilot version loads AGENTS.md natively and picks up .github/copilot-instructions.md. The instructions file is the guaranteed path; AGENTS.md native support has been rolling out and varies by version.
• That .github/prompts/ slash prompts are enabled in your Copilot (the prompt-files feature). If your version names or locates them differently, the bodies are thin so they're trivial to move.
• That Copilot's agent mode can run the .mjs skill scripts (terminal tool enabled).

Your Code App authoring skill

You mentioned adding your own skill to steer how Copilot writes Code App code. When you do, drop it in .agents/skills/<name>/SKILL.md and .claude/skills/<name>/SKILL.md (keep both identical), then add a matching thin .github/prompts/<name>.prompt.md. Say the word and I'll scaffold that empty slot, wired into all three, for you to fill.

🤖 Generated with Claude Code

[Copilot]

Pull request overview

Adds GitHub Copilot–native repo guidance so the existing “harness” (steering doc + skills) is discoverable and runnable when the demo is executed in Copilot rather than Claude.

Changes:

• Add .github/copilot-instructions.md to auto-load the project steering guidance (points to AGENTS.md, restates key “laws” and the npm run verify gate).
• Add four Copilot / prompt files under .github/prompts/ that act as thin entry points to the canonical .agents/skills/*/SKILL.md docs (no duplicated skill logic).
• Encode the intended run order and task entry convention (“Complete Task N”) to keep Copilot aligned with the existing task workflow.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
.github/copilot-instructions.md	Repo-level Copilot instructions pointing at AGENTS.md and summarizing the harness rules + workflow.
.github/prompts/scaffold-code-app.prompt.md	Adds /scaffold-code-app prompt that routes to the existing scaffold skill and enforces the quality-tooling step.
.github/prompts/dataverse-scaffold-tables.prompt.md	Adds /dataverse-scaffold-tables prompt pointing to the scaffold-tables skill and the schema file.
.github/prompts/dataverse-create-records.prompt.md	Adds /dataverse-create-records prompt pointing to the seed skill and the seed YAML file.
.github/prompts/dataverse-add-data-source.prompt.md	Adds /dataverse-add-data-source prompt pointing to the add-data-source skill and the correct table logical names/order.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.