Microsoft Fabric UX / deterministic auditing / component health

Checklist Auditor

Checklist Auditor is a deterministic component-health system for the Microsoft Fabric UX repo. It turns the team's component review checklist into executable rules that scan Web Component source, options files, Storybook stories, README documentation, wrapper surfaces, generated metadata, and package exports.

Its purpose is simple: catch platform drift before product teams inherit it. In a component platform with more than 120 components across Web Components, Angular wrappers, and React wrappers, quality problems rarely show up as one obvious failure. They show up as small inconsistencies spread across the surfaces that define a component.

5,000+ bugs, inconsistencies, missing exports, documentation gaps, and metadata issues resolved
120+ Microsoft Fabric UX components covered by repeatable component-health checks
3 implementation surfaces kept aligned: Web Components, Angular, and React
Rules component review conventions converted into deterministic repairs

Audit pipeline

The checklist becomes a repeatable repair loop.

Instead of a reviewer manually comparing every artifact, the auditor walks the component contract, identifies drift, applies deterministic fixes where possible, and leaves reviewable findings where judgment is needed.

Scan

Read every surface

Source, options, exports, wrappers, Storybook, docs, metadata, and release files.

Compare

Detect drift

Find missing APIs, stale docs, invalid controls, wrapper gaps, and inventory mismatches.

Repair

Apply safe fixes

Regenerate docs, add exports, normalize options, and refresh generated contracts.

Review

Surface judgment calls

Escalate ambiguous component behavior, unsupported examples, or API decisions.

Protect

Keep migration clean

Ship a catalog that humans and agents can consume without rediscovering hidden drift.

What it solved

Component-platform drift is rarely dramatic. A component can exist in source but be missing from a barrel export. A React wrapper can expose less than the Web Component supports. A Storybook story can teach an unsupported composition. A README can omit a public API member. An options file can use a different naming convention. Generated metadata can quietly stop matching implementation.

Checklist Auditor makes those problems visible and fixable at scale. The first major audit pass resolved more than 5,000 bugs, inconsistencies, missing exports, documentation gaps, metadata gaps, wrapper mismatches, and checklist violations across the component catalog.

Why it matters

The auditor protects the core Microsoft Fabric UX operating model: Web Components as the source of truth, with thin Angular and React wrappers that stay aligned. Without tooling, that model is fragile. Every new component or API change creates several chances for drift across implementation, docs, stories, wrappers, exports, contracts, and release metadata.

Instead of asking reviewers to remember every convention across every component, the repo can encode those conventions as rules. Instead of letting product teams discover missing exports or wrapper gaps during migration, the platform catches them earlier.

What it checks

Consistency across every surface that defines a component.

Source

Public APIs, attributes, properties, events, slots, parts, CSS custom properties, lifecycle patterns, and implementation conventions.

Options

Exported option arrays, normalized values, naming conventions, type definitions, and options metadata.

Exports

Missing component exports, missing options exports, incorrect index wiring, package surface drift, and barrel files.

Storybook

Missing argTypes, unsupported controls, stale examples, missing state coverage, and stories that drift from real APIs.

Docs

Missing API sections, stale tables, undocumented members, missing accessibility notes, and incomplete usage guidance.

Wrappers

Missing wrapper props, event mismatches, wrapper API drift, and framework adapters that differ from the source component.

Metadata

Stale custom-elements metadata, inconsistent component inventory, missing coverage records, and outdated contracts.

Release

Missing changelog or change-file coverage, generated inventory gaps, and package changes that lack release evidence.

Kinds of bugs it found

  • Components implemented in source but not exported correctly.
  • Options defined in code but missing from stories, docs, or wrappers.
  • Storybook controls that allowed invalid values or omitted supported values.
  • README API tables that were stale or incomplete.
  • Wrapper props that did not match the underlying Web Component.
  • Event names, attribute names, option names, slots, parts, and CSS properties that drifted across files.

Deterministic fixes

The key design choice was that many issues should not require an LLM or a reviewer to think about them every time. If a rule can determine the right fix, the auditor can apply it deterministically.

That includes repairs like adding missing exports, regenerating README sections, updating Storybook argTypes, normalizing option metadata, refreshing contract files, and aligning generated component inventories. It is more than a report generator; it is a repair tool.

Agentic workflow role

Checklist Auditor grounds the repo's agentic workflows. Agents can explain findings, route fixes, summarize component gaps, or generate follow-up plans, but the auditor keeps the truth layer deterministic.

It gives agents structured findings instead of vague review comments, and it gives reviewers a stable way to verify that component changes preserve the platform contract. AI accelerates the workflow; Checklist Auditor protects correctness.

Platform leverage

The result is component health as an engineering system instead of a manual review ritual. Every encoded convention removes future review work, makes migrations less surprising, and keeps the Microsoft Fabric UX component platform aligned as it scales.

Fewer missing imports, fewer broken examples, fewer undocumented APIs, fewer wrapper surprises, and fewer repeated comments means more of the platform's energy goes into product capability instead of rediscovering drift.