AWS Cloudscape design system report
Cloudscape shows useful public design-system readiness, while a separate purposive agent benchmark found uneven contract compliance.
How well does Cloudscape stay aligned from design through production?
68.75/ 100
Cloudscape scored 68.75 out of 100 for design-through-production readiness, a mixed result. Foundations, Components, and Structure each scored 75, showing substantial public support. Bindings scored 56.25 and Governance scored 62.5, so teams still need clearer design-to-code mappings and stronger published proof that lifecycle checks run continuously. This readiness score is separate from the agent benchmark below.
Evidence boundary: This report distinguishes missing public evidence from a missing capability. When a current result was not found in public sources, AWS may still have it internally and can share it for a re-run.
Foundations
75 / 100
Cloudscape’s public token model is strong; complete public proof of release-wide parity and drift control was not found.
Bindings
56.25 / 100
Design and code assets are available, but the evaluated public evidence did not show a complete maintained mapping across them.
Components
75 / 100
Cloudscape has broad component coverage and detailed public quality guidance, with incomplete public proof across every required state and environment.
Structure
75 / 100
Cloudscape documents application layout and product patterns well; unified end-to-end coverage evidence was not found publicly.
Governance
62.5 / 100
Cloudscape publishes contribution, ownership, and release practices; public operating trends and decisions were not demonstrated.
FoundationsThe evaluated sources document semantic tokens, modes, density, and delivery. They did not provide one current public result reconciling every Figma variable, released token output, mode, and approved exception.Strong shared values help AWS products remain consistent across themes, densities, and implementation contexts.Publish executable token-selection examples with stable source anchors and every required response field. Also confirm whether AWS compares Figma variables with every released token output and density mode; if not, add a release check that blocks unexplained differences.BindingsThe study found useful Figma and code surfaces alongside sampled Button, Modal, density, and token differences. It did not find one versioned public mapping that reconciles Figma properties, design variables, released tokens, and typed component APIs.Unexplained differences force designers, engineers, and agents to guess which surface is authoritative.Publish copyable, versioned typed examples for nested navigation, required fields, and Modal composition alongside one mapping across Figma properties, design variables, released tokens, typed APIs, and approved exceptions. Fail a release when a sampled difference has no owner and rationale.ComponentsThe public sources show mature components plus accessibility and responsive guidance. The study did not find one current public coverage result spanning all supported states, keyboard paths, screen sizes, and assistive-technology expectations.Teams need to know that components work in the difficult states and environments users actually encounter.Publish exact TypeScript examples and negative examples for stateful component APIs, then show the required states, screen sizes, keyboard paths, and accessibility checks for each supported component with current results.StructureThe evaluated guidance covers layout, validation, loading, empty states, and creation patterns. It did not provide one current public result showing complete flows passing together across focus, loading, recovery, accessibility, and narrow screens.Products fail where components meet data, navigation, focus, loading, and recovery, even when each component works by itself.Publish executable filtered-list, load-more, and create-resource references using supported imports and immutable typed state. Cover validation, loading, success, failure, retry, live announcements, and narrow screens, then require three new runs to pass every runtime check.GovernanceThe public repositories describe issues, reviews, continuous integration, ownership, and release evidence. The study did not find public trend data connecting adoption, migrations, exceptions, review speed, or recurring quality results to operating decisions.Leaders need to see whether changes move safely, migrations finish, and recurring quality problems close.Set targets for reviews, migrations, adoption, exceptions, and regressions. Publish the trend, owner, and follow-up decision when a target is missed.What did the recorded agent meet in the Cloudscape benchmark?
The same 20 purposively selected tasks were run using the recorded repetition design under two operational profiles. A run counted as a pass only when it met every blocking answer, evidence, and harness requirement under the corrected v2 evaluator.
Most runs missed at least one benchmark requirement. This is not a product-quality score.
The difference is descriptive because the profiles changed more than documentation access.
What this comparison cannot show: The selected tasks are not a representative sample of production work, and the operational profiles differed in tool availability, command flags, output mode, parser, and preamble. The result does not estimate production reliability or the causal effect of documentation access.
Results by what the task actually measured
| Measurement | Evidence supplied with each task | Optional official Cloudscape resources |
|---|---|---|
| Executable software | 11 of 36 | 8 of 36 |
| Contract transcription | 10 of 18 | 10 of 18 |
| Semantic selection | 0 of 3 | 1 of 3 |
| Governance procedure | 3 of 3 | 3 of 3 |
Software-producing tasks, exact-answer transcription, semantic selection, and governance reasoning are reported separately because they are different kinds of evidence.
Foundations
Evidence supplied with each task: 6 of 12 · Optional official Cloudscape resources: 6 of 12
Choose the AppLayout background token for light and dark mode
Identify Cloudscape’s semantic main-content background token in JavaScript and Sass, support both visual modes, avoid hard-coded colors, and cite the governing guidance.
Choose the token for a visible keyboard focus ring
Select Cloudscape’s semantic focus-border token in JavaScript and Sass, avoid raw color values, and cite the token source supplied with the task.
Use Cloudscape’s responsive spacing system
Choose Cloudscape’s element-query strategy and scaled large-spacing token without inventing pixel values or media queries.
Style a custom surface without hard-coded colors
Pair semantic body-text and main-background tokens for light and dark mode, and avoid direct CSS overrides of Cloudscape components.
Bindings
Evidence supplied with each task: 7 of 12 · Optional official Cloudscape resources: 3 of 12
Connect single-row Table selection to React state
Build a typed controlled Table with stable row identity, named selection controls, and a browser-rendered selection update.
Connect a controlled Modal to its open and close actions
Build a typed Modal with an explicit header, footer actions, controlled visibility, and working open and Cancel behavior.
Update SideNavigation after an internal link is followed
Keep activeHref in React state, prevent full-page navigation, and update the active item from the Cloudscape follow event.
Connect a required name field to validation state
Build a typed FormField and Input pair whose error clears after the user enters a resource name.
Components
Evidence supplied with each task: 4 of 12 · Optional official Cloudscape resources: 2 of 12
Build a keyboard-operable automatic-updates toggle
Use controlled Toggle state, update it from the Cloudscape change event, announce the preference, and keep it usable at 320 pixels.
Build a retryable and dismissible error message
Use Flashbar and Button to show a load error, expose a retry state, support dismissal, pass accessibility checks, and fit a narrow viewport.
Make both Cancel and Escape close a Modal
Open a controlled Modal, close it through the Cancel action and onDismiss path, pass accessibility checks, and fit a narrow viewport.
Show loading, disable Save, then show Saved
Use a Cloudscape Button to expose idle, pending, and completed states while preventing duplicate submission.
Structure
Evidence supplied with each task: 0 of 12 · Optional official Cloudscape resources: 3 of 12
Build empty, populated, and filtered inventory states
Compose AppLayout, Table, Input, and Button so an empty collection can create a resource and a filtered no-match state does not show the create action.
Build create, configure, save, success, and retry states
Validate and focus a required name, configure an option, save the reviewed payload, show progress, and recover from a rejected save.
Filter a resource table and recover from no matches
Filter Alpha and Beta rows, show a distinct no-match state, and restore the full inventory when the filter is cleared.
Preserve rows, show a Spinner, and append loaded results
Keep Alpha visible while loading, render a visible Cloudscape Spinner beside the status, then add Beta without prescribing a timer implementation.
Governance
Evidence supplied with each task: 7 of 12 · Optional official Cloudscape resources: 8 of 12
Plan the public evidence needed to release a change
Order the documented test, visual, bundle-size, ownership, and release-note evidence, and leave undocumented deployment approval explicitly unknown.
Plan an accessibility regression fix from report to release
Order the public issue, focused accessibility and visual tests, reviewed change, continuous-integration checks, and release evidence.
Plan the review path for an accessibility fix
Document the defect, add a focused accessibility test, pass review and continuous integration, and retain release evidence.
Use public repository evidence to plan a reviewed component change
Construct a review sequence from public build, test, ownership, and contribution evidence without inventing an internal approval step.
Evidence requirements were a separate source of failure
The profile using only the evidence supplied with each task satisfied every required source and reviewed citation region in 41 of 60 runs. The optional-resource profile did so in 36 of 60 runs. These checks measure evidence compliance, not executable correctness.
The optional official-resource tool was actually used in 12 of 60 runs and made 21 requests. The other runs measured tool availability, not exposure to additional documentation.
How Cloudscape can get better
Turn typed APIs and complete workflows into executable agent contracts.
- Problem
- The profile using only the evidence supplied with each task passed 24 of 60 benchmark runs. Executable failures included typed API and workflow-state problems; other failures involved exact semantic, governance, or evidence requirements.
- Action
- Publish versioned, copyable TypeScript references for SideNavigation items, Modal dismissal, Flashbar messages, filtered lists, load-more behavior, and the complete create-resource flow. Pair them with stable evidence anchors and tests for type, accessibility, interaction, responsive behavior, semantic selection, and provenance.
- Owner
- Cloudscape component, accessibility, pattern, documentation, and developer-experience teams
- Finish line
- Each reference passes its applicable type, browser, accessibility, responsive, semantic, and evidence checks. A preregistered rerun reports each measurement kind separately and applies its preregistered decision rule.
- Save for the next evaluation
- Published references, negative fixtures, task contracts, source snapshots, runtime versions, test outputs, actor streams, and the preregistered analysis plan
- Decision enabled
- Which public examples and acceptance checks reduce failures in this benchmark, without treating the result as production reliability.
Keep Figma and released code aligned before changes ship.
- Problem
- This study found unreconciled Button and Modal controls, density modes, token inventories, and no single maintained Figma-to-code mapping. AWS may already track some of these differences internally.
- Action
- Generate a versioned manifest of Figma variables and properties, released tokens, typed component APIs, mappings, and approved exceptions. Compare it in release CI and block unexplained differences.
- Owner
- Cloudscape foundations, Figma library, component, documentation, and build-infrastructure teams
- Finish line
- A seeded mismatch fails the release check and identifies the exact token, property, mapping, or output involved.
- Save for the next evaluation
- Cross-surface manifest, CI results, mapping coverage, compatibility report, drift trend, and approved exceptions
- Decision enabled
- Whether a Figma, token, mapping, or component API change is safe to release.
Test complete product experiences across the whole system.
- Problem
- This study did not find one published coverage view spanning required component states, screen sizes, accessibility checks, layout rules, and end-to-end product flows. AWS may maintain broader internal results.
- Action
- Define the required states and environments for each supported component and flow. Run the matrix in CI and publish current pass, failure, and exception results.
- Owner
- Cloudscape component, pattern, accessibility, product, and quality teams
- Finish line
- Required components, states, flows, and viewports pass with current results and owned, expiring exceptions.
- Save for the next evaluation
- Component quality matrix, flow results, accessibility output, viewport coverage, and exception register
- Decision enabled
- Whether a supported component or product flow is ready across Cloudscape’s promised conditions.
Use operating results to decide where Cloudscape needs investment.
- Problem
- This study did not find public trend data for adoption, migrations, exceptions, review speed, or quality. AWS may track these internally; sharing current results would allow them to be assessed in a re-run.
- Action
- For each operating review, publish the target, current result, trend, accountable owner, and next action for reviews, migrations, adoption, exceptions, and regressions. Record the staffing or process decision made when a target is missed.
- Owner
- Cloudscape governance, release, program, and design-infrastructure leadership
- Finish line
- The team can show current trends and the decisions those trends changed.
- Save for the next evaluation
- Metrics pack with baselines, targets, trends, owners, and follow-up decisions
- Decision enabled
- Where leadership should invest, delay a migration, or change staffing, ownership, or process.
The bottom line
Publish versioned TypeScript examples for typed collections, navigation, Modal dismissal, Flashbar recovery, and complete creation flows. Give decision-relevant guidance stable anchors, and require generated work to pass the applicable type, interaction, accessibility, responsive, semantic, and evidence checks. Save the examples, negative fixtures, test output, evaluator version, and source snapshots for a preregistered rerun.
Access the evidence
Review the scoring method, criterion-level rationales, citations, and downloadable study files.