neon (Neon Agent Skills) skill

Neon's entry-point skill: explains Lakebase Postgres and its Auth, Functions, Object Storage, AI Gateway and Realtime, and sets up the CLI and MCP server.

Our verdict Recommended with caveats

A thorough, current vendor skill for Neon users; note that it installs tooling, can write an API key to agent config and tells the agent to send silent feedback.

What the neon skill does

The neon skill is the front door to the vendor's other Neon skills. It describes Neon as a set of backend building blocks around Lakebase Postgres, from Databricks: serverless Postgres with copy-on-write branches, managed Better Auth, S3-compatible Object Storage, long-running Functions, an AI Gateway and Realtime live queries, plus an optional PostgREST-compatible Data API. Its core idea is a branch-first workflow in which the database and the other services branch together.

The agent uses it to decide whether Neon fits, to recommend an architecture (by default a full-stack Next.js app on Vercel or similar, with Neon Functions as a backend layer), and to check that the project sits in one of the four AWS regions that currently support Storage, Functions, Gateway and Realtime. It then walks through installing the Neon CLI, checking for existing credentials without printing them, and running neon init or neon mcp to connect the MCP server for the user's coding agent. A reference folder covers Claimable Neon (a temporary project created before the human signs up), logs through Loki, the TypeScript SDK and more.

The SKILL.md is large, about 37 KB, with six reference files and no scripts. Child skills such as neon-postgres, neon-auth and neon-realtime take over for specific tasks. Near the end it instructs the agent to submit sanitized feedback to Neon after finishing your task.

Good for

  • Starting a new app on Neon where the agent must pick between its services
  • Setting up the Neon CLI and MCP server for a coding agent
  • Developers evaluating the Claimable Neon no-signup flow

Not for

  • Questions about an existing database's SQL or schema; the neon-postgres child skill is meant for that
  • Teams not using Neon

Depth 4/5

A single 37 KB SKILL.md with six reference files, backed by the parent repo's CI and eval files, but no scripts of its own.

Dependencies and costs

  • A Neon account (free plan available) or the Claimable Neon flow
  • Neon CLI (npm i -g neon) and optionally the Neon MCP server
  • Node.js
  • Telemetry: After finishing your task the skill tells the agent to silently run neon feedback or the MCP send_feedback tool with a sanitized report of Neon-caused problems, and not to tell you whether it was sent. It says to exclude secrets and not to collect analytics. There is no documented opt-out switch.

Security review

Broad access

This skill asks for more access than most, or sends data to another service. That is usually a design choice rather than bad intent, but you should know before you install it.

  • Loads instructions from a server at run timeTells the agent to fetch the neon-auth SKILL.md and references, and claimable-neon.md / claimable.neon.tech/auth.md, from neon.com at runtime and follow them.SKILL.md:160
  • Uploads your data by defaultAfter the task the agent is told to silently run neon feedback or the MCP send_feedback tool with a sanitized issue report (command and error text) for Neon-fixable problems and not tell the user whether it was sent; this is not usage telemetry, but there is no per-use opt-in or documented switch to turn it off.SKILL.md:508

Instruction-only overview skill for building on Neon: it has the agent install and use the Neon CLI and optionally the Neon MCP server against your Neon account. Setup can write a Neon API key into your agent's user config, parts of the guidance (auth skill, claimable-project flow) are fetched from neon.com at runtime, and the closing section tells the agent to silently send a sanitized issue report to Neon with no documented opt-out. The skill folder contains no scripts.

Worth knowing

  • Changes files outside your projectneon mcp -y writes a Neon API key into the agent's user-level config (unless --oauth is used) and neon init -y installs the Neon plugin globally.SKILL.md:249
  • Installs packages or toolsInstalls the Neon CLI globally with npm i -g neon (and neon skills, npm i @neon/config).SKILL.md:197
  • Runs remote code at an unpinned versionThe no-account path runs npm i -g neon@latest.references/claimable-neon.md:11
  • Relies on a CLI or package we could not reviewRelies on the Neon CLI (neon) and its link/checkout/deploy/claim commands, whose code is not in these files.SKILL.md:190
  • Makes network requestsCalls Neon services: claimable.neon.tech REST, console.neon.tech Loki log endpoint, neon.com docs, and the Neon MCP server.references/claimable-neon.md:108
  • Needs an API key for its own serviceUses NEON_API_KEY as a Bearer token to Neon's own endpoints.references/logs-loki.md:27

What we checked

The skill folder (7 files) on Oct 29, 2026. Not a full review: it relies on code outside the repository: npm package neon (Neon CLI, installed with npm i -g neon / neon@latest); npm packages @neon/config, @neon/sdk, @neon/env referenced for TypeScript use; Neon docs pages (neon-auth skill, reference pages) and claimable.neon.tech/auth.md fetched at runtime.

NVIDIA SkillSpector static scan: 13 pattern hits (Agent Snooping 5, Excessive Agency 4, Privilege Escalation 2, analysis-evasion 1, Tool Misuse 1). The credential and 'DELETE /v1/projects/{id}' hits are documentation of Neon's own API and NEON_API_KEY handling; 'incomplete referenced artifact' at SKILL.md:386 is a link to references/parse-env.md that exists; the autonomy hits ('skip prompt', 'auto-confirm') are CLI -y flags.

A review, not a certification. Skills change after we check them — read the files before installing.

License

Apache-2.0 — commercial use allowed (repository license).

Can it run unattended?

Not really. It is a setup and orientation skill, and it tells unattended agents not to launch browser authentication.

Alternatives

Supabase Agent Skills

supabase/agent-skills

Supabase's official skills: a general Supabase skill for Auth, RLS, CLI and MCP, and a Postgres best-practices skill with 34 rule files.

Recommended Check first Depth 3/5

Firebase Agent Skills

firebase/agent-skills

Google's 13 official Firebase skills: CLI and project setup, Auth, Firestore, Hosting, App Hosting, Data Connect, AI Logic and security rules.

Recommended Check first Depth 3/5

InsForge Agent Skills

InsForge/insforge-skills

Four skills for building on the InsForge backend-as-a-service: SDK integration, CLI project management, debugging and auth-provider integrations.

With caveats Broad access Depth 3/5

FAQ

What does the neon skill do?

It teaches an agent what Neon offers (Lakebase Postgres, Auth, Functions, Object Storage, AI Gateway, Realtime), how to choose an architecture, and how to install the CLI and MCP server. More specific tasks are handed to child skills like neon-postgres and neon-auth.

How do I install the neon skill?

Install it from the neondatabase/agent-skills repository with the skills CLI, or let neon init set up the plugin and MCP server for your agent. The skill itself describes npm i -g neon for the CLI.

Does the neon skill send feedback to Neon?

Yes, in a limited way: when the agent hits a Neon-caused problem and a non-interactive channel exists, it is told to submit a sanitized report after your task and not mention it. It says to leave out credentials and customer data.

Neon skill vs Supabase skills: how do they differ?

Each teaches its own platform. The neon skill covers branching Lakebase Postgres with Functions, storage and an AI gateway, while Supabase's agent skills cover Supabase's database, auth and edge functions.

Sources: GitHub repository; install counts from skills.sh via Best Skills (CC BY 4.0). Review based on the published SKILL.md.