Choose checks for cmux changes, add tests that reproduce observable bugs, and verify that Swift test files are included in build targets. It also guides local and CI validation, including evidence that selected tests actually ran.
Choose scoped cmux verification, add behavioral tests, and validate Swift test targets and wiring. Use when adding tests or deciding what local/CI evidence a change needs.
Review the instructions the agent will receive before you install.
Run repository commands only from a trusted checkout;
even verify-local.py --help and --list load repository code.
| Task | Command |
|---|---|
| Choose checks and parse changed Swift | python3 scripts/verify-local.py |
| Run the full CI static recipe | python3 scripts/verify-local.py --all |
| Parse current Swift edits | python3 scripts/verify-local.py --only swift-syntax --swift-changed |
| Check new Swift test-file wiring | python3 scripts/verify-local.py --only test-wiring |
| See what a UI test did, one frame per action | scripts/ui-test ClassName or scripts/ui-test <run URL> (guide) |
| Render view code to PNGs in seconds, light and dark, without building the app | scripts/ui-lab/ui-lab.py <harness> [--watch] (guide) |
| Dogfood the app from CI: drive it with a JSON tour and get screenshots and accessibility trees | scripts/run-e2e.sh --scenario dogfood/scenarios/<tour>.json --ref <sha> --frames (guide) |
| Read the screenshots and GIF CI posted of an app PR's build before merging it | the PR's dogfood comment (guide) |
| Decide whether a red web test is the PR's fault or a flake | python3 scripts/ci/web_test_verdict.py --pr <N> (guide; --coverage runs the PR's code) |
Add a base ref after --swift-changed to include committed changes. Use --list
to find other checks and --help for options. Parsing checks syntax; it doesn't
typecheck or run tests.
Read the command guide for piped paths or JSON receipts. Use the validation guide to choose package, native, web or runtime checks. Docs and portable tooling use their scoped checks.
Keep a focused command that fails on the reported symptom, then rerun it after the repair. Setup failures and zero executed tests don't demonstrate the bug. Exercise one behavior at a time so a failure identifies what needs fixing.
Keep two commits: first the failing behavioral regression, then the fix. Run the same focused command on both and record the commit SHAs, the expected failure and the passing result. A setup failure or zero executed tests is not regression proof. When the proof is available locally, push both commits together after the fix passes; a separate hosted CI run on the deliberately broken intermediate commit is unnecessary. If the failure only reproduces in CI, use that lane and keep its receipts. Required CI and review still apply to the final pushed head.
New cmuxTests/*.swift files need both PBXFileReference and Sources build-phase
membership in cmux.xcodeproj/project.pbxproj. Add through Xcode or follow a wired
sibling, then run the wiring check above: an unwired file can otherwise produce
a misleading zero-test pass.
After creating, renaming, or deleting a direct cmuxTests/*.swift file, run ./scripts/sync-test-wiring. It deterministically reconciles the PBXFileReference, PBXBuildFile, cmuxTests group child, and cmuxTests Sources membership; --check performs the same validation without writing. Foreign target membership is rejected with an explicit diagnostic. New Sources/**/*.swift app files are wired with ./scripts/wire-app-sources.py (--check lists unwired ones); UI tests with --target cmuxUITests --dir cmuxUITests; run it after any merge that took main's project.pbxproj, which drops a branch's app-source entries. The workflow-guard-tests CI job still runs ./scripts/lint-pbxproj-test-wiring.sh as a defensive Sources-phase guard.
Swift unit/integration targets use Swift Testing (import Testing, @Test,
@Suite, #expect, #require). Portable Python/shell guards retain their existing
frameworks. UI tests remain XCTest/XCUITest; do not migrate XCUIApplication tests.
New Swift package test targets start on Swift Testing. Prefer parameterized tests
for repeated cases and tags for selection. Use .serialized for suites that
require ordering, not locks or sleeps. Migrate an existing XCTest file only when
an edit already crosses it; see the migration mapping.
An app build does not compile test targets. Package/refactor and public API changes need the relevant test target compiled, then the selected tests actually executed. Follow build-for-testing and execution guidance; report skipped/unsupported checks explicitly.
For remote tmux sizing changes, use the E2E recipe.
Normal PR CI already runs the suites a diff edits or touches. full-ci and
unit-ci are not review or merge requirements; see
PR CI coverage before adding either.
No reviews yet. Tell others what you thought.
No login needed. Share how it worked for you, ask a question, or note where it did not run.
No comments yet. Be the first to write one.
Organize cmux windows, workspaces, and split panes so an agent can target the right terminal reliably. Create terminals with startup commands, move or reorder panels, change focus, and flash a panel to identify it.
Guide substantial Swift changes in cmux with rules for package boundaries, dependency direction, state management, concurrency, and API documentation. Structure code so public package types can be tested without launching the app.
Guide cmux backend changes with consistent rules for Effect, a TypeScript library for structured error handling. Covers Cloud VM state, database migrations, provider integrations, and authentication and payment checks so changes follow the intended architecture.
Maintain cmux billing code with guidance on Stripe checkout, pricing, subscriptions, and access rights. It maps the relevant services and provides workflows for testing purchases, handling payment events, and rolling out billing changes.
Automate browsing in cmux panels without taking focus away from current work. Target a specific panel to open pages, fill forms, click controls, wait for page changes, extract content, and inspect download history.
Capture cmux windows as screenshots or video clips from the CLI. Create visual evidence for pull requests and bug reports, with region selection and captions to explain drags, animations, or layout changes.