agentsclimarketplace

Vercel platform security

Skill hlsitechio/claude-skills-security/appsec-stack-pack/vercel-platform-security

Security audit for applications deployed on Vercel covering environment variable scoping (Production/Preview/Development), Deployment Protection, Edge Config secrets, Vercel Cron auth, Image Optimization SSRF, custom headers via vercel.json, branch/deployment URL exposure, and Vercel-specific platform concerns. Use this skill whenever the user mentions Vercel, vercel.json, vercel deploy, Edge Config, Vercel Cron, Deployment Protection, preview deployments, or asks "audit my Vercel deployment", "Vercel security review". Trigger when the codebase contains `vercel.json`, `.vercel/`, or Vercel is the deployment target.From its SKILL.md

Install
npx -y skills add hlsitechio/claude-skills-security --skill vercel-platform-security

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

7.7 KB, ~1.8k tokens by cl100k_base, as published. Nobody here has run it

Vercel Platform Security Audit

Audit the Vercel deployment configuration. Application-level concerns covered in framework skills; this skill is about Vercel-specific surface.

When this skill applies

  • Reviewing vercel.json configuration
  • Auditing environment variable scoping across Production / Preview / Development
  • Reviewing Deployment Protection settings
  • Checking Vercel Cron and webhook setups
  • Auditing Edge Config and Edge Network usage

Workflow

Follow ../_shared/audit-workflow.md.

Phase 1: Stack detection

ls vercel.json .vercel/ 2>/dev/null
# Vercel CLI
vercel --version 2>/dev/null

Phase 2: Inventory

cat vercel.json 2>/dev/null
ls -la .env* 2>/dev/null
# Cron config
grep -nE 'crons:' vercel.json 2>/dev/null

Phase 3: Detection — the checks

Environment variables

Vercel scopes env vars to Production / Preview / Development.

  • VRC-ENV-1 Production secrets NOT replicated to Preview. Preview deployments are accessible to anyone with the URL (unless Deployment Protection is on); preview env having production DB credentials = breach.
  • VRC-ENV-2 Preview deployments either use a separate (preview) database OR have access controls that prevent leaking.
  • VRC-ENV-3 Sensitive variables marked "Sensitive" in Vercel UI (mask the value from team members without specific perms).
  • VRC-ENV-4 Variables prefixed NEXT_PUBLIC_ / VITE_ / etc. truly public — see framework-specific skills.
  • VRC-ENV-5 No env vars in vercel.json (they get committed). Use the dashboard or vercel env.

Deployment Protection

By default, Vercel makes every preview URL public. For private apps, this is data exposure.

  • VRC-DP-1 Deployment Protection enabled for Preview deployments. Options:
    • Standard Protection: only team members can access preview URLs (recommended).
    • Password Protection: shared password (lower-friction sharing with non-team).
    • Vercel Authentication: SSO via Vercel.
  • VRC-DP-2 Production also protected for staff-only apps (admin dashboards, internal tools).
  • VRC-DP-3 Trusted IPs / bypass tokens reviewed; not shared widely.
  • VRC-DP-4 OPTIONS preflight allowed through protection so APIs work cross-origin from protected URLs.

Branch / deployment URLs

  • VRC-URL-1 Branch URLs (<project>-git-<branch>-<team>.vercel.app) and commit URLs known to leak in PRs, logs, monitoring. Treat all as semi-public.
  • VRC-URL-2 Custom domains for production; alias-only deployments not used for sensitive data.
  • VRC-URL-3 robots.txt prevents indexing of preview URLs (Vercel default does this; verify if customized).

vercel.json configuration

Headers, rewrites, redirects, cron all live here:

  • VRC-CFG-1 Custom headers (CSP, HSTS, etc.) defined in headers section for static assets that aren't handled by the framework's middleware:
    {
      "headers": [
        { "source": "/(.*)",
          "headers": [
            { "key": "Strict-Transport-Security", "value": "max-age=63072000; includeSubDomains; preload" },
            { "key": "X-Content-Type-Options", "value": "nosniff" }
          ]
        }
      ]
    }
    
  • VRC-CFG-2 Rewrites don't accept user-controllable destinations (open redirect class via rewrite source patterns).
  • VRC-CFG-3 Redirects with source: "/(.*)/" patterns vetted.

Vercel Cron

{
  "crons": [{
    "path": "/api/cron/cleanup",
    "schedule": "0 0 * * *"
  }]
}
  • VRC-CR-1 Cron endpoints verify the request comes from Vercel — use Authorization: Bearer ${CRON_SECRET} header. Vercel sends this when you set the CRON_SECRET env var.
  • VRC-CR-2 Without CRON_SECRET, ANY HTTP request to /api/cron/cleanup would trigger the job — attacker can run it arbitrarily.
  • VRC-CR-3 Cron endpoints idempotent; multiple invocations don't double-process.
// app/api/cron/cleanup/route.ts
export const GET = async (req: Request) => {
  if (req.headers.get('authorization') !== `Bearer ${process.env.CRON_SECRET}`) {
    return new Response('Unauthorized', { status: 401 });
  }
  // ... job
};

Edge Config

Edge Config is a fast read-only store. Treat data as semi-public (low-sensitivity).

  • VRC-EC-1 Edge Config token (EDGE_CONFIG) is a secret — keep server-only.
  • VRC-EC-2 Data in Edge Config is feature flags, configuration — not secrets or per-user data.
  • VRC-EC-3 Writes to Edge Config go through Vercel API with separate auth; not exposed to runtime code.

Image Optimization (next/image, etc.)

  • VRC-IMG-1 images.remotePatterns (Next.js) specific; see nextjs-security.
  • VRC-IMG-2 Vercel's edge image service blocks RFC1918 IPs for SSRF; verified by default.

Functions runtime

  • VRC-FN-1 Serverless Functions: cold-start state not leaking across invocations (don't write per-request data to module-level variables).
  • VRC-FN-2 Edge Functions: secrets accessed via process.env are bound at deploy time. Don't compute secrets dynamically based on user input.
  • VRC-FN-3 Function regions appropriate for data residency requirements.

Logs

  • VRC-LOG-1 Vercel Log Drains configured for production observability; logs centralized.
  • VRC-LOG-2 Logs don't include full request bodies with PII / secrets — sanitize before logging.
  • VRC-LOG-3 Vercel's default log retention is short; long-term storage configured separately.

Webhooks

  • VRC-WH-1 Vercel deployment webhooks (sent to your endpoints on deploy events) verified via signature.
  • VRC-WH-2 GitHub / GitLab integration webhooks (used internally by Vercel) — verify the GitHub App permissions if you've adjusted defaults.

Project membership

  • VRC-PM-1 Vercel team access reviewed periodically. Stale developer access removed.
  • VRC-PM-2 Production deployment permissions limited (require approval).
  • VRC-PM-3 Integrations (Slack, GitHub, etc.) scope minimized.

Custom domains and DNS

  • VRC-DNS-1 Custom domain SSL auto-renewed.
  • VRC-DNS-2 DNS records use CNAME to Vercel; not direct A records that lock the domain to a specific IP (less flexible).
  • VRC-DNS-3 Subdomain takeover risk: removed projects with custom subdomain pointers cleaned up.

Build configuration

  • VRC-BLD-1 Build command and install command match local; no overrides allowing arbitrary commands.
  • VRC-BLD-2 Build cache scope appropriate; don't share build cache across unrelated projects.
  • VRC-BLD-3 Environment variables exposed during build (vs runtime) reviewed — build-time vars baked into static assets are effectively public.

Analytics and Speed Insights

  • VRC-AN-1 Vercel Analytics enabled doesn't conflict with privacy regulations (GDPR consent flow, etc.).
  • VRC-AN-2 Speed Insights script (@vercel/speed-insights) doesn't capture sensitive URL paths in metrics.

Phase 4: Triage

Critical: production secrets in Preview env; Cron endpoint without CRON_SECRET check; Deployment Protection off for sensitive Preview URLs.

Phase 5: Report

Use ../_shared/findings-schema.md. Prefix IDs with VRC-.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Gives 1 of the 12 instructions most security skills give in ~1.8k tokens

Counted across 648 of the 828 authors here whose files we hold, read 2026-08-07

  • Parameterize all database queriesin 68 of 648, across 51 files
  • Hash passwords using bcrypt, scrypt, or argon2in 49 of 648, across 36 files
  • Apply rate limiting to authentication endpointsin 48 of 648, across 24 files
  • Configure security headershere, and in 35 of 648, across 19 files
  • Validate all inputsin 32 of 648, across 24 files
  • Validate all external input at the system boundaryin 29 of 648, across 19 files
  • Run containers as a non-root userin 28 of 648, across 15 files
  • Use httponly secure samesite cookies for sessionsin 26 of 648, across 15 files
  • Run dependency audits before every releasein 21 of 648, across 10 files
  • Encode output to prevent cross-site scriptingin 21 of 648, across 11 files
  • Copy dependencies before source codein 20 of 648, across 9 files
  • Store secrets in environment variablesin 20 of 648, across 18 files

Said here and by no other author read

  • scope production secrets away from preview deployments
  • enable deployment protection on preview deployments
  • verify all cron requests using a secret token
  • store only non-sensitive data in edge config
  • prevent cross-invocation state leaks in functions
  • prevent open redirects in rewrite rules

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 326,871. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.