How do I install the Ekx Vercel Deploy skill?

Run npx skillmds@latest add ekinoxis-evm/ekx-vercel-deploy in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Ekx Vercel Deploy safe to use?

This skill has not completed SkillMD's automated safety review yet. Capability flags: reads secrets. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Ekx Vercel Deploy?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Ekx Vercel Deploy free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Ekx Vercel Deploy?

Ekinoxis-evm (@ekinoxis-evm) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Ekx Vercel Deploy by ekinoxis-evm — Agent Skill

Name: Ekx Vercel Deploy
Rating: 4.2 (5 reviews)
Author: Ekinoxis-evm

Ekx Vercel Deploy

Deploying and operating Ekinoxis apps on Vercel — project linking, environment variables per environment, preview vs production deploys, build failures, runtime logs and errors, cron jobs, and domain setup. Use when a deploy fails, when an app works locally but not in production, when adding or rotating an env var, when setting up a scheduled job, or when configuring a custom domain.

Ekinoxis-evm Updated Sep 22, 2026

File contents

Vercel

Every web surface we ship. 18 projects.

Authoritative: the vercel MCP server (in 5 of our repos) — use it to read build logs, runtime errors and deployment status rather than asking the user to paste them. Plus the bundled vercel-cli and deployments-cicd skills.

CLI

vercel link                       # connect the local folder to a project
vercel env pull .env.local        # pull dev env vars down — do this first, always
vercel                            # preview deploy
vercel --prod                     # production
vercel logs <url>                 # runtime logs
vercel inspect <url>              # build output + timings

vercel env pull is the correct way to get a working .env.local. Copying from a teammate's laptop is how the three different Supabase key names got into the portfolio.

Environments

Three scopes: Development, Preview, Production. A variable set only for Production is absent in preview — which is why preview builds fail with "supabaseUrl is required" while production is fine.

Set per environment deliberately:

Test/sandbox keys (Stripe sk_test_, Base Sepolia RPC) → Development + Preview
Live keys → Production only

NEXT_PUBLIC_* changes need a redeploy, not just a save — they are baked in at build time.

Vercel provides for free: VERCEL_ENV, VERCEL_URL, VERCEL_GIT_COMMIT_SHA, VERCEL_PROJECT_PRODUCTION_URL. Use VERCEL_ENV to branch behaviour rather than inventing your own flag.

const isProd = process.env.VERCEL_ENV === "production";

Preview deploys

Every push gets a URL. The trap: third-party allowlists do not know about it.

Add the wildcard preview domain to:

Privy dashboard → allowed domains (*.vercel.app or the project's preview pattern)
Alchemy → allowed referrers
Supabase → Auth → redirect URLs
Stripe → webhook endpoint (or use the CLI locally instead)

A login that works in production and silently fails in preview is almost always this.

Cron

// vercel.json
{ "crons": [{ "path": "/api/cron/scan", "schedule": "*/15 * * * *" }] }

Cron routes are publicly reachable URLs. Protect them:

if (req.headers.get("authorization") !== `Bearer ${process.env.CRON_SECRET}`) {
  return new Response("no", { status: 401 });
}

Vercel sends that header automatically when CRON_SECRET is set. Limits: Hobby is one cron/day; Pro allows minute-level. A 15-minute scanner needs Pro.

Functions have a max duration (10s Hobby, 60s+ Pro, configurable). A long scan must either stream, chunk across invocations, or move to Railway — which is exactly why our long-running P2P agent runs on Railway, not Vercel.

Build failures — the usual four

Type errors. Vercel type-checks; next dev does not. Run npx tsc --noEmit before pushing.
Missing env var in that environment. supabaseUrl is required = the var is not set for Preview.
Case-sensitive imports. macOS is case-insensitive, the build container is not. ./Button vs ./button builds locally and fails on Vercel.
Lockfile drift. Commit the lockfile; the build uses it exactly.

Gotchas

Env var set for the wrong environment — the #1 preview failure.
NEXT_PUBLIC_ without redeploy stays stale.
Preview URLs missing from third-party allowlists.
Unprotected cron routes are public endpoints.
Function timeout on long jobs.
Case-sensitive filesystem in the build container.
.vercel/ is gitignored — it holds the project link, and is per-developer.

---name: ekx-vercel-deploydescription: Deploying and operating Ekinoxis apps on Vercel — project linking, environment variables per environment, preview vs production deploys, build failures, runtime logs and errors, cron jobs, and domain setup. Use when a deploy fails, when an app works locally but not in production, when adding or rotating an env var, when setting up a scheduled job, or when configuring a custom domain.---# VercelEvery web surface we ship. 18 projects.Authoritative: the **`vercel` MCP server** (in 5 of our repos) — use it to read buildlogs, runtime errors and deployment status rather than asking the user to paste them.Plus the bundled `vercel-cli` and `deployments-cicd` skills.---## CLI```bashvercel link # connect the local folder to a projectvercel env pull .env.local # pull dev env vars down — do this first, alwaysvercel # preview deployvercel --prod # productionvercel logs <url> # runtime logsvercel inspect <url> # build output + timings````vercel env pull` is the correct way to get a working `.env.local`. Copying from ateammate's laptop is how the three different Supabase key names got into the portfolio.---## EnvironmentsThree scopes: **Development**, **Preview**, **Production**. A variable set only forProduction is *absent* in preview — which is why preview builds fail with"supabaseUrl is required" while production is fine.Set per environment deliberately:- Test/sandbox keys (Stripe `sk_test_`, Base Sepolia RPC) → Development + Preview- Live keys → Production only**`NEXT_PUBLIC_*` changes need a redeploy**, not just a save — they are baked in atbuild time.Vercel provides for free: `VERCEL_ENV`, `VERCEL_URL`, `VERCEL_GIT_COMMIT_SHA`,`VERCEL_PROJECT_PRODUCTION_URL`. Use `VERCEL_ENV` to branch behaviour rather thaninventing your own flag.```tsconst isProd = process.env.VERCEL_ENV === "production";```---## Preview deploysEvery push gets a URL. The trap: **third-party allowlists do not know about it.**Add the wildcard preview domain to:- **Privy** dashboard → allowed domains (`*.vercel.app` or the project's preview pattern)- **Alchemy** → allowed referrers- **Supabase** → Auth → redirect URLs- **Stripe** → webhook endpoint (or use the CLI locally instead)A login that works in production and silently fails in preview is almost always this.---## Cron```json// vercel.json{ "crons": [{ "path": "/api/cron/scan", "schedule": "*/15 * * * *" }] }```Cron routes are **publicly reachable URLs**. Protect them:```tsif (req.headers.get("authorization") !== `Bearer ${process.env.CRON_SECRET}`) { return new Response("no", { status: 401 });}```Vercel sends that header automatically when `CRON_SECRET` is set. Limits: Hobby isone cron/day; Pro allows minute-level. A 15-minute scanner needs Pro.Functions have a **max duration** (10s Hobby, 60s+ Pro, configurable). A long scanmust either stream, chunk across invocations, or move to Railway — which is exactlywhy our long-running P2P agent runs on Railway, not Vercel.---## Build failures — the usual four1. **Type errors.** Vercel type-checks; `next dev` does not. Run `npx tsc --noEmit` before pushing.2. **Missing env var** in that environment. `supabaseUrl is required` = the var is not set for Preview.3. **Case-sensitive imports.** macOS is case-insensitive, the build container is not. `./Button` vs `./button` builds locally and fails on Vercel.4. **Lockfile drift.** Commit the lockfile; the build uses it exactly.---## Gotchas1. **Env var set for the wrong environment** — the #1 preview failure.2. **`NEXT_PUBLIC_` without redeploy** stays stale.3. **Preview URLs missing from third-party allowlists.**4. **Unprotected cron routes** are public endpoints.5. **Function timeout** on long jobs.6. **Case-sensitive filesystem** in the build container.7. **`.vercel/` is gitignored** — it holds the project link, and is per-developer.

Frequently asked questions