groundwork 0.3.2 (beta)

0.3.2

New features

• Reviews can run chosen reviewers on another model through a command-line tool installed on your machine: the GitHub Copilot CLI or the Codex CLI. Describe the model in plain words under review.external in GROUNDWORK.md, for example writing: "latest high-end Gemini". At each review, groundwork picks a matching installed tool and model, keeps that model read-only, and labels its findings with the tool and model it used. Its findings are verified like any other reviewer's. When no installed model fits or the tool fails, the reviewer is reported as not run with a fix to try; it is never silently replaced. So far this has been tried with the writing reviewer.
• Onboarding's review step asks who runs the reviewers. The agent you are using stays the recommended choice; an external model is offered as an advanced option.
• When a plugin update brings more than 8 new choices, update lists them and sends you to onboarding to set them up, instead of asking about each one.
• Onboarding's issue tracker step explains why keeping issue records helps.

Bug fixes

• The banner shown when onboarding finishes renders correctly.

groundwork 0.3.1 (beta)

0.3.1

New features

• Journey usability tests now show how easy each journey felt, not only whether it passed: an overall ease rating, the effort at each milestone, and the hardest step. The full report is saved to a file, and the conversation shows a short plain-language summary with a link to it.

groundwork 0.3.0 (beta)

0.3.0

New features

• Journey usability tests (add-on): an AI participant that has never seen the code walks a whole journey on the running app, such as landing page → sign-up → organization setup, and reports where it hesitates, gets lost, or gives up. Participants act as typical users with assigned habits and slips, and the first run proposes journeys for you to confirm. It can also gate every release.
• /groundwork:update applies the changes from a new plugin release to an existing setup, asking only about what changed instead of rerunning onboarding.

groundwork 0.2.0 (beta)

0.2.0

New features

• After onboarding, optionally set up extra workflows: Slack work reports, session handoff notes, weekly summaries, release notes, and scheduled release preparation. Browse them later with /groundwork:addons.
• Code-based UX reviews are more precise: findings follow a clear severity scale, problems that existed before the change are listed separately, layout risks that need the running app are handed off instead of reported as findings, and bugs and wording are left to the reviewers that own them.

groundwork 0.1.2 (beta)

0.1.2

New features

• Choose separate schedules for UX reviews and UI QA, and review usability from the code or in the running app.
• See a clear completion message after onboarding successfully saves your configuration.

Upgrading

• Run onboarding again to migrate existing screen-check settings and choose how UX reviews run.
• QA frequency now controls UI QA only; UX reviews have their own frequency and method. Existing configurations keep their previous UX triggers until onboarding migrates them. If you previously used both code-based and on-app UX checks, onboarding asks which method to keep.

groundwork 0.1.1 (beta)

0.1.1

New features

• Start a tracked change directly from a request, with planning when the work needs it.
• Choose manual, pre-release, or continuous UI and UX checks, and adjust review effort per task.
• Receive Claude Code update notices for future releases.

Bug fixes

• Finish onboarding with grouped questions, fewer repeated prompts, and a concise settings summary.
• Keep parallel tasks isolated in worktrees and focus later review rounds on fixes.

Upgrading

• Run onboarding again to refresh existing repository rules and migrate legacy flight workflows.
• Existing 0.1.0 installations need one manual update before future update notices become available.

groundwork 0.1.0 (beta)

First beta of groundwork, a Claude Code and Codex plugin that sets up how an AI coding agent works in your repository and then runs that workflow.

Beta. The skills come from workflows used daily in two real projects, but this packaged form has not run end to end yet. Expect rough edges, and please open an issue when something does not behave as the README says.

Install

/plugin marketplace add enebin/groundwork
/plugin install groundwork@groundwork

Then run /groundwork:onboarding in your repository.

What's in 0.1.0

• Onboarding — eight questions (language, issue tracker, workflow, branches, verification, architecture, review, anything else). Every choice is a bundle compared by pros, cons, and token cost, with a recommended option first. Writes GROUNDWORK.md and the chosen rules into AGENTS.md / ARCHITECTURE.md only after you approve.
• Workflows — step-loop (plan, step), design-first, tdd-loop, flight (boarding → landing, plus rejoin, release, autoland), quick (review-and-fix, commit-push), review-only.
• Review — review runs architecture, code-quality, security, writing, and SDK-usability reviewers in parallel; ux-review runs one usability agent per persona; ui-qa checks the running UI against the task and design source. Works with skeptic-review when installed.
• Issue trackers — local files (default), GitHub Projects, Linear, Jira.
• Claude Code and Codex — one set of skills, two manifests.

Least tested

• The Jira tracker, the tdd-loop workflow, the visual-regression add-on, and shared local issues (shared: true).
• Token cost levels are relative estimates, not measurements.