Who can start a run
- From the dashboard or the API, only a signed-in member of your organization, or an API key made in it, can start a run. Viewers cannot.
- Our GitHub App acts only on repositories you installed it on. Through the App, a pull request on a private repository is checked only when it carries the shipguarde label or a collaborator asks for a run in a comment that mentions @shipguarde.
- Every message from GitHub is checked against its signature before we act on it.
- Before a run starts, we check its target address. Private, loopback and cloud metadata addresses are refused.
- During a run, every connection the browser makes is checked the same way, including redirects, frames and background requests.
Your logins
- The agent only logs in to your site if you give it a login, either for the project or for a single run.
- A saved project login is encrypted with AES-256-GCM before it is stored.
- The planning model is never given the password. It leaves the password field empty, and our runtime types the password in.
- The screenshot after the step that types a password is not saved.
- A saved project login is never used on a pull request from a fork.
What the agent can do in a browser
- It can click, type, hover, drag, scroll, press a key, pick from a dropdown, go to a URL, wait, and check the screen. It has no action to upload or download a file.
- Around steps that submit, pay, order, send or delete, a failed step only counts as recovered when the checking model also rules out that the action happened twice.
- Each run has limits: a time budget (3 minutes by default), a timeout per step (45 seconds by default), at most 100 steps, at most 3 re-plans, and a model cost ceiling of $1 per browser run in production.
- Each run starts a fresh browser with no cookies or saved state from other runs, and the browser is closed when the run ends.
Your repository and GitHub
- You choose which repositories our GitHub App can see when you install it.
- The App asks GitHub for read access to code, deployments and metadata, and write access to checks, commit statuses, pull requests and issues. The write access is used to post the check result, the commit status and comments on the pull request.
- Code is read with short-lived GitHub tokens. When your code is cloned, the token can only read that one repository.
- Lint, secret scanning and dependency checks run in a separate service that holds none of our database, model or GitHub App keys.
What you can check after a run
Every step records a screenshot before and after (except the after shot of a password step), the reasoning of the model that acted, and the reasoning of the model that checked the result. All of it is on the run page, so you can see why a step passed or failed.
Your code and pages sent to models
- Every model call is restricted, in code, to zero-retention endpoints that do not train on what we send.
- We do not capture model prompts in production.
Where your data is kept
- ShipGuarde runs on Railway, in Singapore: the app, the database, the browser runs, and the service that runs lint, secret scanning and dependency checks.
- The dependency check looks up the package names and versions in your npm lockfile on npm's public registry, in the United States. It sends no source code.
- Screenshots are stored in a private Cloudflare R2 bucket. They are served only through our API, to members of the organization that owns the run or through a share link a member made.
- A pull request run cannot be shared by link. A URL run report becomes public only when a member makes a share link, and turning the link off stops it working at once.
- Runs, findings and screenshots are kept until you delete them. An admin can delete a project, which deletes its runs, findings and run screenshots.
Sign-in and access
- Sign-in is handled by Clerk. Each person in your organization is an owner, admin, member or viewer. Deleting a project and managing API keys, billing and GitHub settings need an admin or owner.
- Runs, findings and screenshots are only returned to members of the organization that owns them, apart from a URL run a member has shared by link.
- An API key itself is stored only as a SHA-256 hash.
- Traffic to shipguarde.com and our API uses HTTPS. Plain HTTP is redirected, and the site sends a Strict-Transport-Security header.
Page checks and the free scan
- The accessibility, performance, SEO tag, console and network, and cookie checks load your page once and read it.
- The free scan on our home page, which needs no account, runs only read-only page checks. It does not click, type, fill a form or log in.
- The free scan refuses login, account and checkout paths.
- The free scan is limited to 3 scans an hour and 10 a day per IP address, and 100 a day in total.
Report a vulnerability
If you find a security issue, please tell us privately at [email protected] before you share it publicly. We will reply, look into it, and keep you updated on what we find.