Conciseness

Every line delivers non-obvious, Stream-specific knowledge Claude cannot already know ("Don't write Stream SDK code from memory. APIs change between majors", "CreateBlockList 400s if the list already exists - check ListBlockLists first", "Never use a useRef flag as a run-once guard in an effect with cleanup"). Not 4: no section explains known concepts or could be trimmed without losing actionable content.

Actionability

Copy-paste-ready commands and exact identifiers throughout: "cat $(getstream docs chat-sdk/react)/INDEX.md", "getstream api UpdateApp --request '{\"video_primary_use_case\":\"livestreaming\"}'", package names ("stream-chat stream-chat-react", "@stream-io/video-react-sdk"), hooks ("useCreateChatClient()"), and concrete checks ("Check: send **bold**, a list, and a code block"). Not 4: guidance is executable end-to-end, down to error codes and fallbacks ("Fallback: render pills from message.reaction_groups").

Workflow Clarity

The build flow is sequenced Setup → Auth → Building → Verify with explicit validation: "Type-check and build, run the dev server, and test multi-user flows in two tabs as two different users", "Run the check for each default in Defaults to keep", and the Defaults section supplies per-feature "Check:" checklist items. Error-recovery guidance exists ("CreateBlockList 400s if the list already exists - check ListBlockLists first"), and the upgrade/audit workflows are each explicitly sequenced. Not 4: checkpoints and checklists are explicit rather than implicit.

Progressive Disclosure

The body is a concise overview with uniformly well-signaled one-level-deep references for each deep topic: "follow chat-versions.md", "Follow stream-agents.md", "follow design-matching.md", "follow sendbird-migration.md", "follow ai-support-agent.md", "ai-integration.md". Note: no bundle files are present in this review copy, so the referenced files cannot be verified on disk, but the structure itself fully matches the anchor — clear overview, appropriate split, easy navigation. Not 4: no bulk detail is inlined that belongs in a reference file.

Specificity

"Build, integrate, audit, or upgrade Stream Chat, Video, Feeds, and Moderation" plus "Scaffold a new app, add Stream to an existing project, review an integration, upgrade an SDK major, or migrate from Sendbird" lists multiple concrete actions with comprehensive coverage across the product surface. Not 4: coverage goes beyond 'several actions with minor gaps' — it enumerates both the operations and the domains (Chat, Video, Feeds, Moderation, AI assistants, agents).

Completeness

The 'what' is explicit ("Build, integrate, audit, or upgrade Stream Chat, Video, Feeds, and Moderation in React / Next.js web apps") and a 'when' is present via "the default skill for web work" and "Web React only - React Native uses stream-react-native". Not 5: there is no explicit 'Use when...' clause with concrete trigger phrases, so the 'when' could be more specific; not 3: the default-skill statement plus the React Native boundary constitutes explicit-enough trigger guidance.

Trigger Term Quality

Natural user vocabulary is comprehensively covered: "team messaging, DMs, video conferencing, livestreaming, social feeds, AI assistants", "React / Next.js", "Sendbird", "React Native" — terms users would actually say when needing this skill. Not 4: synonyms are covered across all product areas (messaging/DMs/team chat, video/livestreaming) plus the competitor name and platform markers.

Distinctiveness Conflict Risk

"Web React only - React Native uses stream-react-native" explicitly disambiguates the sibling skill, and named products (Stream Chat/Video/Feeds/Moderation, "@stream-io/stream-agents") plus the Sendbird migration niche carve a clear, minimal-conflict trigger space. Not 4: overlap risk with other skills is essentially nil.