Privacy
Last updated 21 September 2026
Firms upload Office Actions, which are privileged client material. This page says plainly what happens to them. It describes the software's actual behaviour rather than the widest permission a policy could claim.
What we collect
Documents you upload. Office Actions, and the prior papers you choose to attach to a matter. We extract structured fields from them — the rejections, the claims, the examiner, the art unit — and generate a response document.
Account information. Your name, email address, the firm you belong to and your role in it. If you sign in with Microsoft, Google or Apple, we receive your email address and name from them and nothing else.
Usage records. Which documents were generated, by whom and when; an audit record of downloads; and the response balance ledger. Firms ask who downloaded a file during security review, and that cannot be reconstructed after the fact, so it is recorded from the start.
Payment details are never ours. Card numbers go directly to Stripe and do not pass through our servers. We store a customer id and a subscription id.
What we do not do
- We do not train models on your documents. Not ours, and we do not permit our providers to. No content you upload is used to improve any model.
- We do not log document contents. Application logs and error reports carry record identifiers and error codes. Extracted fields, document text and client reference numbers are excluded by rule, and our error reporting strips request bodies before anything is sent.
- We do not sell or share data with advertisers. There is no advertising, no analytics broker and no third-party tracking on the application.
- We do not read your matters. Access by us for support requires a specific request from you.
Who else processes it
Delivering the product requires the services below. Each one is limited to what its function requires.
| Service | What it does | What it sees |
|---|---|---|
| Supabase | Database, authentication, file storage | Everything stored, encrypted at rest |
| Anthropic | Reads the Office Action and extracts its fields | The document, for the duration of the request |
| Google Cloud Document AI | Text recognition, and only for scanned documents with no text layer | The document, for the duration of the request |
| Vercel | Hosting and request routing | Requests in transit |
| Stripe | Payments and subscriptions | Billing contact and payment details; no documents |
| Mailgun | Account email — invitations, password resets | Email addresses and message text; no documents |
| PostHog | Visit analytics for this website, without cookies | Pages viewed, links clicked and session recordings on useskeleton.com, with form inputs masked; never the application or any document |
| Sentry | Error reporting | Stack traces and identifiers, configured to exclude document content |
Where the product retrieves a cited reference, it requests that document from a public patent source. The request contains the reference's own publication number and nothing about you, your firm or your matter.
How long we keep it
Your firm sets its own retention period, from one day to ten years. When it is set, each uploaded document is stamped with a purge date on arrival and deleted by a scheduled job when that date passes. A firm that sets no period keeps its documents until it deletes them — indefinite retention is available, but it has to be chosen rather than arrived at.
Billing records and the audit log outlive the documents they refer to, because they have to: an audit record that is purged with its document cannot answer the question it exists to answer.
Who can see what, inside your firm
Every record belongs to a firm, and that boundary is enforced by the database itself rather than by application code — a query that forgot to filter returns nothing rather than someone else's matter. Uploaded files are stored in private buckets under your firm's identifier.
Within a firm, every member can see every matter. There is no per-matter access control today. Firms that need conflict screens should know this before they sign up; it is on the roadmap and it is not built.
Your requests
Write to support@useskeleton.com to get a copy of your firm's data, correct it, or have it deleted. An owner can delete the firm's account, which removes its documents and its stored extractions.
Data is stored and processed in the United States. If you are subject to the GDPR or the UK GDPR, we act as processor for the documents you upload and as controller for your account record, and a data processing agreement is available on request.
Changes
Material changes will be sent to account owners by email before they take effect, not posted quietly with a new date at the top.