You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Make desktop behavior and package-cost claims for the agentic component library
provable. Contributors and agents should use one supported desktop-driver and
Storybook workflow, CI results must distinguish native evidence from simulated
evidence, and component/theming bundle boundaries must be reproducible.
Status
In progress. Four of nine direct work items are closed; five remain open.
Added the public @fluentui-react-native/desktop-driver package with a
client-neutral W3C remote end, deterministic fake host, typed client,
WebdriverIO runner, JSON CLI, bounded agent API, and confined artifacts. This
closes #4258.
Moved reusable Storybook desktop orchestration into the shared storybook-desktop CLI and removed app-local lifecycle scripts and the
dedicated windows-tests harness.
Added pull-request smoke jobs for Windows Fabric and Win32 Paper with artifact
upload. The merged validation traversed 136 Windows stories and 130 Win32
stories; three authored plans passed through the fake target on each.
Added initial Button, Checkbox, and Input role/state/value plans.
The bundle-size warning pipeline
(#4253),
Win32 Paper endpoint
(#4259), and
initial desktop Storybook pipelines
(#4279) were
already closed. PR #4295
also added cross-platform theme bundle scenarios that prove modern Flex paths
exclude reverse legacy compatibility payloads.
Current evidence boundary
stories traverses the supported catalog in a real native Storybook app. stories-and-tests performs that native traversal and then runs authored plans
against the manifest-derived fake target. A green authored-plan result is
therefore not yet proof that native accessibility or interaction behavior
passed.
What is left
Add real native driver providers and pass the release gate on macOS, Windows
Fabric, and Win32 Paper
(#4257).
Run authored Storybook plans through those native providers, then define
required jobs, retry limits, quarantine, artifact retention, and ownership
(#4256).
Expand the accessibility schema beyond the initial role/state/value cohort
to normalized accessible name, keyboard reachability, reports, and real
macOS/Windows execution
(#4255).
Finish the Windows smoke-harness disposition by proving that all required
focus-crash coverage survives the retired harness
(#4254).
Evaluate and deliver reliable window-scoped Windows Composition screenshots
with locked-session diagnostics and provenance
(#4260).
The public desktop-driver package is merged, built by the unified graph,
and covered by package tests.
Storybook exposes one shared CLI for manifest generation, native catalog
traversal, fake authored-plan execution, and owned process lifecycles.
Pull-request smoke jobs exercise Windows Fabric and Win32 Paper and upload
artifacts.
The Win32 Paper Storybook endpoint bundles and launches through the shared
workflow.
Cross-platform component and Flex-theme bundle baselines and dependency
boundaries are checked and reported.
At least one authored plan executes against a real native provider with
truthful application-termination and artifact reporting.
At least three components beyond Button have portable role, name, state,
value, and keyboard assertions validated on a real desktop endpoint.
Every false-green path in the driver release task is fixed or documented
as an explicit limitation.
The replacement suite preserves all required focus-crash coverage from
the retired Windows harness.
Production exit criteria
macOS, Windows Fabric, and Win32 Paper run the same authored manifest as
required CI gates with matching digests.
Retry, quarantine, expiry, artifact retention/redaction, and escalation
policies have named owners.
Backend prerequisites and compatibility ranges are published and tested.
The driver package is released and validated by an external consumer
without a globally installed Appium driver.
Bundle growth budgets are required gates and baseline changes require
reviewed rationale.
Flex-only consumers exclude legacy and optional capabilities unless those
capabilities are explicitly imported.
Windows Fabric publishes window-scoped visual evidence with capture
provenance and locked-session diagnostics; pixel differences remain
evidence, not a pass/fail gate.
Portable accessibility assertions are required for the initial component
cohort on macOS and Windows.
Agent context
Architectural boundary
The supported public contract is driver-client neutral and Appium-free.
Platform providers may use platform-specific automation internally, but that
detail must not leak into portable authored plans.
A fake-host pass proves schema and orchestration determinism only.
Native story traversal proves the application can enumerate and render the
supported catalog, not that authored accessibility assertions ran natively.
Deterministic gates and optional visual evidence must remain separate.
Current catalog exclusions
Windows Fabric excludes Accordion and Callout because they still fail-fast
the RNW 0.81 Fabric host during traversal.
Win32 Paper excludes Accordion and ListItem; Callout remains covered on the
Paper endpoint.
These exclusions must remain explicit in manifests and reports until the
underlying platform failures are resolved.
Sequencing
Native providers and false-green fixes precede required interactive jobs.
Accessibility normalization should stabilize before expanding the authored
cohort.
Real Windows sessions should evaluate native screenshot fidelity before a
host-level fallback is built.
Required bundle budgets should use observed variance rather than thresholds
selected merely to make the first run pass.
macOS automation needs an authenticated, unlocked GUI session and stable
Automation Mode; Windows native runs need an unlocked interactive desktop.
Goal
Make desktop behavior and package-cost claims for the agentic component library
provable. Contributors and agents should use one supported desktop-driver and
Storybook workflow, CI results must distinguish native evidence from simulated
evidence, and component/theming bundle boundaries must be reproducible.
Status
In progress. Four of nine direct work items are closed; five remain open.
PR #4294
landed the shared Stage 1 foundation:
@fluentui-react-native/desktop-driverpackage with aclient-neutral W3C remote end, deterministic fake host, typed client,
WebdriverIO runner, JSON CLI, bounded agent API, and confined artifacts. This
closes #4258.
storybook-desktopCLI and removed app-local lifecycle scripts and thededicated
windows-testsharness.Storybook/Metro/driver lifecycles, and
smoke --mode stories|stories-and-tests.upload. The merged validation traversed 136 Windows stories and 130 Win32
stories; three authored plans passed through the fake target on each.
The bundle-size warning pipeline
(#4253),
Win32 Paper endpoint
(#4259), and
initial desktop Storybook pipelines
(#4279) were
already closed. PR #4295
also added cross-platform theme bundle scenarios that prove modern Flex paths
exclude reverse legacy compatibility payloads.
Current evidence boundary
storiestraverses the supported catalog in a real native Storybook app.stories-and-testsperforms that native traversal and then runs authored plansagainst the manifest-derived fake target. A green authored-plan result is
therefore not yet proof that native accessibility or interaction behavior
passed.
What is left
Fabric, and Win32 Paper
(#4257).
required jobs, retry limits, quarantine, artifact retention, and ownership
(#4256).
to normalized accessible name, keyboard reachability, reports, and real
macOS/Windows execution
(#4255).
focus-crash coverage survives the retired harness
(#4254).
with locked-session diagnostics and provenance
(#4260).
budgets after normal variance is established; add a follow-up issue if that
work is no longer represented by closed Infrastructure: Add bundle-size and dependency regression gates #4253.
Work items
Beta exit criteria
and covered by package tests.
traversal, fake authored-plan execution, and owned process lifecycles.
artifacts.
workflow.
boundaries are checked and reported.
truthful application-termination and artifact reporting.
value, and keyboard assertions validated on a real desktop endpoint.
as an explicit limitation.
the retired Windows harness.
Production exit criteria
required CI gates with matching digests.
policies have named owners.
without a globally installed Appium driver.
reviewed rationale.
capabilities are explicitly imported.
provenance and locked-session diagnostics; pixel differences remain
evidence, not a pass/fail gate.
cohort on macOS and Windows.
Agent context
Architectural boundary
detail must not leak into portable authored plans.
supported catalog, not that authored accessibility assertions ran natively.
Current catalog exclusions
the RNW 0.81 Fabric host during traversal.
Paper endpoint.
These exclusions must remain explicit in manifests and reports until the
underlying platform failures are resolved.
Sequencing
cohort.
host-level fallback is built.
selected merely to make the first run pass.
Automation Mode; Windows native runs need an unlocked interactive desktop.
Merged evidence
@fluentui-react-native/desktop-driverstorybook-desktopCLI and generated platform manifests