storybook

Storybook evidence profile

Type: System profile
Status: Active
Canonical for: Factual interpretation of preserved Storybook manifests, MCP and testing evidence
Depends on: corpus/storybook/

Role in the field

Storybook turns component stories and documentation into structured manifests and agent-callable tools. Unlike Figma’s design-file context, Storybook’s primary evidence is executable component behavior, API documentation and tests.

Four-question relevance

• Intent: conditional on authors supplying meaningful JSDoc/MDX.
• Structure: demonstrated for identity, props, descriptions, stories and subcomponents.
• Retrieval: demonstrated through manifests and MCP documentation tools.
• Constraint/evaluation: demonstrated for known story behavior; semantic selection and composition evaluation are not established.

Manifests

• Documented and implemented: Component manifests are generated from CSF plus prop extraction. They include component names, descriptions, source/story paths, import snippets, stories, usage snippets, props, types, required/default values, JSDoc and subcomponents. Manifest documentation
• Documented and implemented: Docs manifests serialize MDX pages, including standalone material such as foundations, accessibility guidance and design tokens.
• Documented: JSDoc is recommended because types alone provide only basic information. The example explicitly distinguishes Button (interaction) from Link (navigation), showing that semantic selection guidance must be authored.
• Implemented: Manifest inclusion is curated through a manifest tag; human-only or instructional stories/docs can be excluded. This is a context boundary, not just extraction.
• Experimental: Manifests and MCP are preview, React-only in the preserved release, and the schema is explicitly unstable.

MCP and agent consumption

• Documented and implemented: The MCP addon exposes three toolsets: development (get-changed-stories, story instructions, preview), documentation (get-documentation, per-story docs, list all), and testing (run-story-tests). MCP API
• Documented: Recommended AGENTS.md instructions tell the agent never to hallucinate properties, query documentation before using a component and run story tests. AI overview
• Implemented: The repository contains the addon, manifest generation, MCP tools and tests. Storybook source snapshot

What becomes machine-readable

Component existence, import path, props, prop types, defaults, examples, documented states, child-component APIs, MDX content and executable story tests. Storybook can also indicate what changed and supply current instructions for creating stories.

What remains implicit

• Choosing between components is only as good as descriptions/JSDoc.
• Stories encode examples, not an exhaustive validity grammar.
• Subcomponents expose parent/child relationships but not broad page/workflow composition.
• Accessibility can be tested for rendered stories, but the manifest does not inherently contain the rationale behind a requirement or guarantee coverage of generated combinations.
• A component absent from the manifest may be intentionally hidden, undocumented or unsupported; agents need policy context to distinguish these cases.

Composition

• Implemented: Stories for multiple components, subcomponent metadata, args, decorators and portable stories provide concrete compositions.
• Inferred: Storybook is a strong representation of known examples and testable states, not a general representation of all good compositions. It can validate a generated story after the fact, but the agent may still lack selection intent before generation.

Governance and quality

• Implemented: Curated manifests control exposure. Story tests provide behavioral verification; existing accessibility, visual and interaction testing can be applied to generated stories.
• Documented: The agent should verify props rather than infer them from conventional names.
• Experimental: The preview/React-only status and unstable schema prevent treating Storybook manifests as an established cross-framework standard.

Evidence assessment

Storybook is the strongest direct implementation of a design system as an agent-queryable component/documentation/test service. It makes APIs and examples highly legible and creates a verification loop. Its central limitation is semantic: structure and examples become queryable, but good product judgment still depends on the quality and explicitness of authored descriptions, rules and patterns.