Checks
Copy MarkdownChecks
Every check has an ID: use it with --only and --skip, in the ignore: list of preflight.yml, and with preflight ignore. preflight checks prints them all. Checks marked opt-in stay off until preflight.yml has their block under checks:. preflight init writes the blocks for secrets and env_parity turned on, and the one for security_headers turned on when you give it a production URL. Asking for an opt-in check that has no block, as in --only secrets, is a usage error (exit 64): no enabled check matches.
SEO and social
| ID | What it checks |
|---|---|
seo_meta |
A title and description, and Open Graph tags |
og_twitter |
og:image, twitter:card, and the rest of the social sharing metadata |
canonical |
A canonical link tag |
structured_data |
JSON-LD Schema.org markup |
viewport |
A viewport meta tag for mobile |
lang |
The lang attribute on <html>, for accessibility |
index_now |
An IndexNow key file, for faster search indexing. When the site answers, the deployed /{key}.txt decides; the source is read only when it can’t be reached. Opt-in |
Security and infrastructure
| ID | What it checks |
|---|---|
security_headers |
HSTS (over HTTPS), Content-Security-Policy, X-Content-Type-Options, and Referrer-Policy, on production and staging. A weak CSP, such as one that allows 'unsafe-inline' scripts, is a suggestion, not a failure. Opt-in (checks.security) |
ssl |
That the SSL certificate is valid, with a warning before it expires. The warning and the error scale with the certificate’s lifetime (an error with under a tenth of it left, a warning under a third), capped at 7 and 30 days, so a short-lived certificate is not flagged the day it is issued. An http:// production URL that redirects to valid HTTPS is a warning. Skipped for a local URL |
www_redirect |
That www and the bare domain redirect to the canonical one |
secrets |
API keys and credentials committed to the code, including credentials in URLs. Every file git tracks is scanned, whatever directory it is in. Files over 2 MB are skipped. Placeholders are not reported, and a key-shaped value in an example file is a warning. When git cannot say which files are tracked, the result is inconclusive. Opt-in (checks.secrets) |
email_auth |
SPF and DMARC DNS records for your sending domain, or the production URL’s domain when none is set (see configuration). Opt-in |
Environment and health
| ID | What it checks |
|---|---|
env_parity |
Variables in .env.example that are missing from .env. Opt-in (checks.envParity) |
health_endpoint |
That the site answers. It tries /health, /healthz, /api/health, /_health, and /status, then the root. With healthEndpoint.path set it tries that path first, and with enabled: true as well, only that path |
Code quality and performance
| ID | What it checks |
|---|---|
vulnerability |
Dependency vulnerabilities, through your package manager’s audit tool, for every project it finds (workspace members covered by a parent lockfile are audited once, and ignore: globs are left out). pnpm and Bun lockfiles are not audited yet: a project with only pnpm-lock.yaml or bun.lock is reported as inconclusive, so run pnpm audit or bun audit yourself |
debug_statements |
console.log, var_dump, debugger, and the like left in the code. A statement behind a development guard is not reported: in code, in Twig, ERB, Blade, Django, and Go templates, and on the same line with && or a ternary |
error_pages |
Custom 404 and 500 pages |
image_optimization |
Images over 500KB |
Legal and compliance
| ID | What it checks |
|---|---|
legal_pages |
A privacy policy and terms of service on the deployed site, or in the source when the host is blocked or down |
Cookie consent tools are checked as services.
Web standard files
| ID | What it checks |
|---|---|
favicon |
A favicon, an apple-touch-icon (.png, .webp, or .svg), and a web manifest, including icons declared by <link> tags in your layout or Next.js icons metadata. A missing favicon is a warning |
robots_txt |
A robots.txt with real rules in it. On production it warns when the file disallows everything for all crawlers or the homepage is marked noindex |
sitemap |
A sitemap.xml, or a generator or route that makes one, including Next.js Pages Router, Astro, and SvelteKit endpoints |
llms_txt |
An llms.txt for AI crawlers |
ads_txt |
An ads.txt, for ad-supported sites. Opt-in |
humans_txt |
A humans.txt crediting the team. Opt-in |
license |
A license file, for open source projects. Opt-in |
For robots_txt, sitemap, llms_txt, ads_txt, and humans_txt, a copy in the repository is not enough when the live site answers and does not serve it: the check warns and says where it looked. When the host blocks the scanner, answers with a server error, or is down, the result is inconclusive, not "not found".
Check IDs renamed in 0.22
Every check ID is now snake_case. 0.22 accepted the old camelCase names with a note on stderr; 0.25 rejects them. An old name in an ignore: list, --only, or --skip is a usage error (exit 64). One error lists every old ID it found, each with its new name and the command that fixes it:
Error: pre-1.0 check IDs are not accepted:
unknown check ID "seoMeta" in ignore: renamed to "seo_meta" in v0.22.0; run 'preflight ignore seo_meta' to update the ignore list
unknown check ID "ogTwitter" in --only: renamed to "og_twitter" in v0.22.0; use --only og_twitterpreflight ignore <new-id> rewrites the old entry in place and removes any duplicate spellings of it. preflight unignore takes either name. The keys under checks: in preflight.yml (healthEndpoint, seoMeta, and so on) are a separate schema and did not change.
| Old | New |
|---|---|
seoMeta |
seo_meta |
ogTwitter |
og_twitter |
securityHeaders |
security_headers |
envParity |
env_parity |
healthEndpoint |
health_endpoint |
robotsTxt |
robots_txt |
llmsTxt |
llms_txt |
adsTxt |
ads_txt |
humansTxt |
humans_txt |
indexNow |
index_now |
Scanning code you don’t trust
Every check reads files; one runs a program. The vulnerability check runs your package manager’s audit tool (npm audit, yarn audit, composer audit, bundle-audit, pip-audit, govulncheck, or cargo-audit) inside the project directory, with a scrubbed environment and a temporary home so it cannot read your tokens. npm’s and Yarn’s user config, cache, and logs are pinned to that temporary home, so a project .npmrc cannot point them back at yours.
It also closes the known ways a repository can hand that run to its own code: Composer runs with --no-plugins --no-scripts, cargo-audit and bundle-audit run directly rather than through cargo and bundle, Composer audits the lockfile with --locked, Yarn classic ignores a yarn-path and every .yarnrc, and Yarn is not run at all at or below a .yarnrc.yml (that target is reported as inconclusive). Git runs with fetching and config-named commands turned off, so a repository’s own git config cannot start a program. Audit tools still read project configuration, so this narrows the risk without removing it. Scanning your own projects is what Preflight is for. When you scan a repository you don’t trust, such as a fork’s pull request in CI, add --skip vulnerability.
To report a security problem in Preflight itself, see its security policy.