AETHER

Aether trust standard

What the badges mean, and what we actually check

Every listing in the Aether catalog carries a trust tier and a signed report. Nothing here is a marketing label: each tier maps to specific evidence, and each scan records the exact checks that produced its score. Reports are HMAC-signed with a versioned key so anyone can re-verify them later.

Trust tiers

Unlisted

Submitted, not yet public.

  • · A draft or pending submission exists
  • · Domain ownership not yet proven
  • · Not discoverable in the public catalog
Listed

Origin verified, automated scan passed.

  • · DNS or HTML domain ownership proof accepted
  • · Every blocking check passed on the live release
  • · Signed trust report published on the listing
Verified

Publisher entity verified on top of the origin.

  • · Everything at Listed
  • · Publisher identity or organisation confirmed
  • · Clean scan history across consecutive releases
Audited

Human review of capabilities and data flows.

  • · Everything at Verified
  • · Manual review of the capability manifest and agent policy
  • · Re-review required on any capability expansion

The checks

Blocking checks must pass for a listing to stay public. Non-blocking checks affect the score and the tier ceiling.

Origin & ownership

Served over HTTPSBlocking

The declared start URL uses https:// and resolves without falling back to plaintext.

Installed web apps hold credentials and tokens. Plaintext transport makes every other guarantee meaningless.

check id: https
Domain ownership verifiedBlocking

A DNS TXT record or a token file at /.well-known on the declared domain proves the publisher controls the origin.

Everything downstream — badges, reports, delisting — is keyed to an origin somebody actually owns.

check id: domain-ownership
Start URL reachableBlocking

The start URL returns a 2xx response within the scan timeout.

Dead listings erode catalog trust faster than bad ones.

check id: reachable

Security posture

Security headersScored

Checks for Content-Security-Policy, Strict-Transport-Security, X-Content-Type-Options, and frame protections.

These headers are the cheapest available defence against injection and clickjacking of an installed surface.

check id: security-headers
Third-party script inventoryScored

Enumerates external script origins loaded by the start URL and flags unexpected trackers.

Publishers should be able to name every origin that executes inside their app.

check id: third-party

PWA installability

Web App ManifestBlocking

A linked manifest with name, start_url, display, and theme_color.

Without it there is no Add to Home Screen, and no store-equivalent install experience.

check id: manifest
Installable icon setScored

Icons at the sizes iOS and Android require for a home-screen launch.

A blurry or missing icon is the single most common reason an install feels untrustworthy.

check id: manifest-icons
Service worker registeredScored

A service worker is registered from the start URL scope.

Offline resilience is the functional parity claim PWAs make against native store apps.

check id: service-worker

Permissions & capabilities

Permissions match declarationBlocking

Sensitive browser APIs detected in the page — geolocation, camera, microphone, clipboard, notifications — are compared against the capabilities the publisher declared.

Undeclared capability use is the failure mode app stores exist to catch.

check id: permissions-declared

Agent & prompt policy

Capability manifest declaredBlocking

Agent listings declare the tools, data sources, and side effects they can invoke.

An agent without a capability manifest cannot be reasoned about by anyone, including its own users.

check id: capability-manifest
Autonomy label justifiedBlocking

The declared autonomy level (assistive, supervised, autonomous) is consistent with the declared capabilities.

An agent that writes to production systems cannot be labelled assistive.

check id: autonomy-label
Data practices disclosedScored

Retention, model-training use, and third-party sharing are disclosed in structured form.

Structured disclosure is what makes catalog-wide compliance filtering possible.

check id: data-practices
Privacy / policy link presentScored

A reachable privacy or usage policy is linked from the app itself.

Users deserve a durable, publisher-owned statement, not just our summary of it.

check id: policy-link

Verifying a report yourself

Every listing detail page exposes its signature and a one-click verification. Signatures are versioned — v2:sha256:… — so reports signed with a retired key still verify, and the response tells you whether the key that signed it is still active.

GET /api/public/verify-report?scanId=<scan-id>

{ "valid": true, "keyId": "v2", "rotated": false,
  "app": "…", "score": 94, "verdict": "pass" }
Browse the catalogRead the full trust standard