Build one small website improvement. Learn when a prompt is enough—and when to turn repeated work into a skill.
In this beginner GitHub Copilot App workshop, you start with a working lab document board. You capture a fictional request, plan a change, implement it, and review a real pull request. Along the way, you author three skills, then package them for sharing.
Full journey: allow 5–6 hours, or two sessions. The suggested schedule is 5 hours 40 minutes including breaks, covering all 13 labs: three learner-authored skills, native cloud review, APM packaging, release, and distribution/configuration review. Only live enterprise-admin execution is optional and role-gated; the distribution lesson is included for everyone.
An optional abbreviated ~3-hour track covers selected hands-on work and previews the later stages. It is not completion of the full journey. No previous skill-authoring experience is needed.
All Caldova people, messages, teams, and documents are original fiction for training. Editorial labels such as “Needs review” are not scientific or clinical judgments. This is not a validated quality-management system, and must not be used for patient care, laboratory operations, or regulated approval. Never put real workplace content or credentials in a public issue or commit.
- Open this repository on GitHub. Choose Use this template → Create a new repository, under your own account or an approved training organization. Do not perform exercises in the shared upstream repository.
- Install/sign in to the GitHub Copilot App. Add your copy in Projects, and start an Interactive project session.
- Follow Lab 00: setup to find that session's actual worktree checkout. Run the following in a terminal in that exact directory, not a different clone.
Cross-platform (macOS, Linux, or PowerShell; Node and npm on PATH):
npm ci
npm run devOpen http://127.0.0.1:5173. The starter shows six fictional records, CDOC-101
through CDOC-106, with no filter. Keep the terminal running. In a second
terminal in the same checkout:
npm test
npm run build
npm run checkThe app uses vanilla JavaScript, HTML, CSS, Vite 7.3.6, and Node's built-in test runner. Use Node 24.21.0 LTS for the rehearsed path (at least that patch within Node 24 LTS) plus Git. No database, container, Azure setup, deployment, website login, or WorkIQ connection is needed to run it. The local server binds to loopback with a strict port; it will fail rather than silently use 5174.
| Starter provides | You create during the workshop |
|---|---|
| Working six-document board and baseline tests | A small view/filter and count, with tests you derive |
| Public fictional source messages | One human-confirmed feature issue in your copy |
| Step-by-step labs and blank, inert worksheets | A bounded plan and short, human-approved spec |
| Packaging helpers and an inert manifest example | intake, plan-to-spec, and code-review skills |
| App CI and a tag-triggered draft-release workflow | A reviewed PR, package manifest/lock, verified archive, and human-published plugin release |
There are no active solution skills or completed feature spec in the starter. Instructor reference files are optional teaching material, not learner context: do not ask Copilot to inspect them. An inert filename prevents automatic registration; it does not prevent an agent from reading that file.
Participants: Git, Node 24.21.0 LTS, a browser, the Copilot App, a GitHub
account with access to your training copy, and a paid Copilot entitlement for
cloud code review. Check AI-credit budgets and applicable App/review policies.
The App itself supports more access options, but those do not necessarily
include cloud review. APM 0.31.0 is introduced only after you author skills;
its native binary needs no Python.
The later packaging lab resolves one pinned public development dependency
with apm lock; that lock-only path does not install its guidance into the App
or export it as runtime content.
Facilitator / administrators: preflight App builds, review eligibility and Actions resources, and arrange the authorized WorkIQ training account, billing, consent, and seeded fictional sources. Repository Admin or an edit repository rules role is needed for the review ruleset. Enterprise ownership is needed only to execute the optional live admin portion of the included distribution lesson. Participants do not provision cloud services. If live prerequisites are unavailable, use explicitly labeled offline practice or a facilitator demonstration and schedule the missing full-journey evidence for later; do not count those substitutes as completed live stages.
- Full-journey schedule and abbreviated track — start here after setup.
- Capability and licensing preflight — dated public references, supported surfaces, and limits.
- Glossary — issue, worktree, skill, PR, package, and policy.
- Facilitator guide — preparation and honest fallback paths.
- Fictional source — the starting request, not a completed issue.
The story is small:
fictional source → issue → plan/spec → PR and tests → reusable skills → plugin
Human review remains at the issue, spec, merge, publication, and administration boundaries. Releasing the skills does not deploy the website.