Conciseness

Mostly efficient and factual, but with recurring waste: the region list appears twice (lines 30 and 58), the caveat that Neon injects only NEON_AI_GATEWAY_* and not OPENAI_* is stated three times (the env-var table, the AI SDK note, and the plain-SDKs section), credential-provisioning instructions are repeated in Setup and again under List Available Models, and promotional prose ("runs on the same Databricks infrastructure that serves trillions of tokens a month", "batteries-included") adds no actionable value. This fits the anchor for mostly efficient content that includes unnecessary explanation and could be tightened, more than the 'minor instances' of the level-4 anchor.

Actionability

Fully executable, copy-paste-ready guidance throughout: the neon.ts config with `neon deploy`, the `neon config status/plan/apply` commands, complete TypeScript examples for @neon/ai-sdk-provider, Mastra, and plain OpenAI SDK, an exact env-var table, the curl for GET /v1/models, and precise dialect paths (${NEON_AI_GATEWAY_BASE_URL}/v1, /openai/v1, /anthropic, /gemini). Common cases (chat completion, streaming, agent with tools, model listing) are all covered with runnable code.

Workflow Clarity

The setup-to-request sequence is clear and mostly checkpointed: check region and plan preconditions, enable aiGateway in neon.ts, run `neon deploy` (with `neon config plan` as a dry-run), confirm env vars, then build the client — plus explicit verification/debug guidance (read /v1/models rather than assuming the catalog, distinguish Free-plan blocking from a reduced catalog, `neon config status`). It falls short of a 5 because the sequence is implied by section order rather than presented as an explicit ordered workflow with feedback loops, and there is no end-to-end 'verify the request works' step.

Progressive Disclosure

The file is a ~290-line monolith with no bundle files: reference-grade detail — the 30-line /v1/models JSON response shape, the plan-gating matrix, per-SDK dialect routing, and full Mastra/agent examples — is inlined in SKILL.md where the rubric expects it split into one-level-deep reference files. Section headers are clear and external pointers (Further Reading, the parent neon skill, models.dev) are well signaled, so it sits at the 'some structure but content that should be separate is inline' anchor rather than the unstructured level-2 anchor.

Specificity

The description lists multiple concrete actions and capabilities: "route between model providers (OpenAI, Anthropic, Google/Gemini, Meta, Alibaba, and more)", "avoid juggling separate provider API keys", and "Works with the OpenAI SDK, Anthropic SDK, google-genai, the Vercel AI SDK, and Mastra by changing only the base URL". Coverage of what the gateway does is comprehensive rather than leaving minor gaps, matching the anchor for multiple specific concrete actions.

Completeness

Both questions are explicitly answered: the "what" ("One API and one credential for frontier and open-source LLMs, built into your Neon branch") and an explicit "Use when..." clause with concrete trigger phrases ("Use when a user wants to call an LLM, add AI/chat/an agent to their app, route between model providers..."). This matches the anchor for clearly and explicitly answering both.

Trigger Term Quality

Trigger phrases are extensive and natural: "call an LLM", "add AI to my app", "chat completion", "model routing", "LLM proxy/gateway", "one API for all models", "use Claude/GPT/Gemini", "AI SDK", "Mastra agent", "Neon AI Gateway", and "log/rate-limit AI calls". Synonyms, product names, and both generic and specific phrasings users would actually say are covered.

Distinctiveness Conflict Risk

The Neon anchoring ("Neon AI Gateway" trigger, "especially when they already use Neon") gives it a clear niche, but several triggers — "call an LLM", "chat completion", "use Claude/GPT/Gemini" — are generic enough to fire for general LLM-SDK requests unrelated to Neon, creating minor-to-moderate overlap with closely related AI skills. Not score 3 because the what-clause firmly ties it to Neon branching and one-credential routing; not score 5 because the generic trigger list does carry real overlap risk.