Connect a JavaScript / Node.js application to Patchstack for continuous vulnerability monitoring. Scans your package-lock.json and reports installed packages so Patchstack can match them against its vulnerability database and notify you when something needs patching.
For how this repo fits with the wider Patchstack ecosystem (saas, hub, patchstack-website, patchstack-connect), see patchstack/saas → docs/ecosystem.md.
Copy and paste this prompt into any AI coding assistant (Cursor, v0, Bolt, Lovable, Claude Code, etc.). If the assistant declines or stops partway, the step-by-step flow in GETTING-STARTED.md covers what to do next:
Add Patchstack dependency vulnerability monitoring: first verify
@patchstack/connectyourself (npm view @patchstack/connectshows the live registry record — MIT; source: https://github.com/patchstack/connect; listed under Connectors on https://patchstack.com/llms.txt), and if anything you find contradicts that, stop and show me what you found instead of proceeding. Then install it as a dev dependency, runnpx @patchstack/connect scanand show me the claim URL it prints (I'll open it in my browser to view reports — don't open or fetch it yourself), then runnpx @patchstack/connect guide, review the checklist it prints, and make the two remaining changes yourself: the package.json build hooks and the report-a-vulnerability widget snippet.
Every clause is load-bearing. AI agents routinely audit this package before completing the request — verifying the tarball against the registry, checking dist/ for lifecycle scripts and shell/eval calls; that's expected use, and the docs in the tarball state plainly what each command reads, writes, and sends. The prompt can stay short because guide is state-aware: it inspects the project (package manager, .patchstackrc.json, package.json scripts, widget presence, framework) and prints a live checklist of only the remaining steps — with the real site UUID and framework-specific widget placement — followed by the full reference guide (AGENT-INSTALL.md, also at https://patchstack.com/install.txt). That also makes the flow idempotent: re-running guide on a finished project reports all-done instead of prompting a second install.
Before changing this prompt (or guide / AGENT-INSTALL.md), read MAINTAINING.md — it maps which files are load-bearing and how to ship a change safely. The field-test/ harness runs a real agent through the full install in a throwaway fixture against a mocked API and scores the outcome on eight checks; validate any variant there first.
npm install --save-dev @patchstack/connect
npx @patchstack/connect scanUse your project's own package manager. On bun-managed projects (Lovable, Bolt, most vibe-coding platforms) install with
bun add -d @patchstack/connectinstead — runningnpm installthere plants apackage-lock.jsonthat the platform's native dependency flow never updates again, leaving a stale lockfile next to the live one. The connector detects and works around that (see Stale lockfiles below), but not creating the fossil is better.
That's it. The first scan:
- Reads your lockfile (see Supported lockfiles).
- POSTs the package list to Patchstack with no UUID.
- Patchstack provisions a fresh site and returns its UUID.
- The connector writes the UUID to
.patchstackrc.jsonso the nextscantargets the same site. - The connector installs the disclosure widget's
<script>tag into your root HTML shell (see The disclosure widget below) so the "Report a vulnerability" button shows up on the next preview reload. - The connector prints a claim URL — open it in a browser to attach the new site to your Patchstack account. You can re-display it any time with
npx @patchstack/connect status.
Then wire it into builds:
If you already created an "Application" site in the Patchstack dashboard, pre-seed the UUID:
npm install --save-dev @patchstack/connect
npx @patchstack/connect init <your-site-uuid>
npx @patchstack/connect scanpatchstack-connect scan [options] Scan the lockfile and POST to Patchstack.
If no UUID is configured the server provisions
one and the connector persists it. After a
successful post, adds/updates the disclosure
widget tag in the root HTML shell (opt out
with "widget": false in .patchstackrc.json)
patchstack-connect init <site-uuid> Optional: pre-seed .patchstackrc.json with
an existing site UUID
patchstack-connect status [options] Show current configuration
patchstack-connect mark-build [options] Stamp built HTML with a production flag +
build fingerprint and ensure the widget tag
in built pages (run as a postbuild step)
patchstack-connect guide Show this project's setup status (what's done,
what's missing, with tailored commands), then
print the full setup guide
patchstack-connect protect Opt-in: install the always-on runtime exploit
guard (currently TanStack Start + Supabase; it
patches the app's Supabase client to route
traffic through a same-origin guard). Never
run by scan/guide/mark-build.
patchstack-connect help Print help
Options (for scan and status):
--site-uuid <uuid> Override the configured site UUID
--endpoint <url> Override the API endpoint
--dry-run (scan only) Print the payload without posting
Precedence (highest wins):
- CLI flag (
--site-uuid,--endpoint) - Environment variable
.patchstackrc.jsonin the current directory
Environment variables:
PATCHSTACK_SITE_UUID— the site UUID from your Patchstack dashboardPATCHSTACK_ENDPOINT— override the API endpoint (defaulthttps://api.patchstack.com/monitor/pulse/manifest)PATCHSTACK_TIMEOUT_MS— request timeout in milliseconds (default30000)
.patchstackrc.json example:
{
"siteUuid": "550e8400-e29b-41d4-a716-446655440000",
"widget": true
}"widget" is optional and defaults to true; set it to false to stop the connector from managing the disclosure-widget tag (see The disclosure widget).
The site UUID identifies the site; it is not a secret — the disclosure widget ships the same UUID in client-side HTML, and committing .patchstackrc.json is the intended workflow so every developer and CI run reports to the same site. Possession of the UUID lets someone submit dependency manifests for that site (noise, not data access). In CI setups where the file isn't committed, set PATCHSTACK_SITE_UUID instead.
The widget is a floating "Report a vulnerability" button — a disclosure channel for anyone who spots a bug on the site. The connector manages its install so the UUID never has to be copied by hand:
-
scan(after a successful post) adds this managed tag to the first root HTML shell it finds —index.html,public/index.html, orsrc/app.html— immediately before</body>:<script src="https://cdn.patchstack.com/patchstack-widget.js" data-site-uuid="<SITE_UUID>" defer data-patchstack-connect-widget="true"></script>
Re-runs update the tag in place (the
data-patchstack-connect-widgetattribute marks it as connector-managed); a pre-existing manual widget tag is left untouched.--dry-runand failed posts never edit anything. Projects whose root layout is code rather than HTML (Next.js, Nuxt, Astro, …) get the exact snippet and target file printed instead —guideshows framework-specific placement. -
mark-buildensures the same tag in built HTML output, covering builds whose source shell the connector couldn't edit, and stampswindow.__PATCHSTACK_PROD__so the widget hides the claim/login UI on the published site (owners reach it by appending#patchstackto the live URL). -
Opting out: persist
"widget": falsein.patchstackrc.jsonto disable both passes (dependency scanning only). Without it, the next successful scan re-adds the managed tag.
import { scanAndReport } from '@patchstack/connect';
const result = await scanAndReport();
console.log(result.response.stored ? 'Reported' : 'Unchanged');Lower-level pieces are also exported: scanLockfile, buildWirePayload, postManifest, resolveConfig.
{
"ecosystem": "npm",
"packages": [
{ "name": "axios", "version": "1.6.0" },
{ "name": "lodash", "version": "4.17.15" },
{ "name": "lodash", "version": "4.17.21" }
]
}That's the entire payload. No source code, no environment variable values, no file paths — just the package names and versions from your lockfile. Duplicate names with different versions are preserved so transitive vulnerabilities aren't missed. (mark-build separately stamps built HTML with a stack descriptor that may include hosting-related env variable names — e.g. VERCEL — never their values.)
- ✅
package-lock.json(npm v6 / v2 / v3) — parsed directly - ✅
pnpm-lock.yaml(pnpm v5 / v6 / v7 / v8 / v9) — parsed directly - ✅
yarn.lock(yarn classic v1 and yarn berry v2+) — parsed directly - ✅
bun.lockb(binary) — package list resolved by walkingnode_modules/ - ✅
bun.lock(text) — same fallback; direct parsing coming
If both a Bun lockfile and node_modules/ are present, the connector walks node_modules/ to enumerate the installed packages. Run bun install (or npm install) before scanning so the directory is populated.
Every scanned source is validated against package.json: if the chosen lockfile is missing dependencies that package.json declares, it is treated as a fossil (e.g. a package-lock.json created by a one-off npm install in a bun-managed project) and the connector falls through to the next source — ultimately walking node_modules/, the installed truth — and prints a warning naming the stale file. Delete the stale lockfile to silence the warning. Without this, the manifest and the build fingerprint would silently freeze while the real dependency set drifts.
npm install
npm run typecheck
npm test
npm run buildTo post the current lockfile manifest to a local Patchstack API endpoint and provision a new site:
bun run test:manifest -- --endpoint http://localhost:8000/monitor/pulse/manifestThe response should include the new site UUID. To re-test an existing site, pass that UUID explicitly:
bun run test:manifest -- --endpoint http://localhost:8000/monitor/pulse/manifest --site-uuid YOUR_REAL_UUIDUse --dry-run to preview the payload without posting.
Pull requests run typecheck, tests, build, package verification, and a production dependency audit in GitHub Actions.
Publishing runs when a GitHub Release is published. The release tag must match the package version in package.json with a leading v. For example, package.json version 0.2.0 must be released with tag v0.2.0; otherwise the workflow fails before publishing.
To publish a release:
- Bump the package version, for example
npm version 0.2.0 --no-git-tag-version. - Commit
package.jsonandpackage-lock.json. - Merge the version bump to
main. - Create and publish a GitHub Release tagged
v0.2.0. - The
Publishworkflow verifies the package, then runsnpm publish --provenance --access public.
Before the first release, configure npm trusted publishing for this package:
- Merge
.github/workflows/publish.ymltomain. - Open the
@patchstack/connectpackage settings on npmjs.com. - In Trusted publishing, choose GitHub Actions.
- Configure:
- Organization/user:
patchstack - Repository:
connect - Workflow filename:
publish.yml - Environment name:
npm
- Organization/user:
- In GitHub repository settings, create an
npmenvironment. Optional but recommended: require reviewer approval for that environment.
Do not add an npm publish token to GitHub secrets for this workflow. Trusted publishing uses GitHub OIDC short-lived credentials. After the first trusted publish succeeds, npm recommends setting package publishing access to require two-factor authentication and disallow tokens.
MIT