Security
This page says how Common Ink keeps your notes and bank data safe, and how to tell us about a vulnerability. A hole in Common Ink would expose someone's private writing, so we take every report seriously.
Effective 7 November 2026.
Reporting a vulnerability
Email it to security@commonink.app. The fix can be worked out with you privately.
Please don't open a public issue, or post about it anywhere else, until a fix is out.
What to include
- What an attacker can do, and to whom: read someone else's note, act as another member, reach another workspace, run script in the app, keep access after it was taken away, and so on.
- The steps to reproduce it, from a fresh account where you can: the requests, the note text, the plugin or embed, the agent token's options.
- Where it happens: the address (
commonink.app, a page,/api/...,/mcp). - What you used: browser and version, MCP client, the
inkCLI. - Anything you saw that belongs to someone else, so it can be dealt with. Don't keep a copy of it.
What to expect
- An acknowledgement that the report was received. We aim to send it within 7 days.
- A first assessment (whether it reproduces, how severe it is, what is in the way of a fix) as soon as one is known, and news as the fix goes on.
- When it is fixed and deployed, the fix is made public, crediting you unless you would rather not be named.
Common Ink is run by one person and has no bug bounty.
In scope
- The app at
commonink.appand every address it answers: the pages, sign-in (/auth/...), the API (/api/...), share links (/s/...), the HTML embed document (/embed-frame), and the MCP endpoint (/mcp). - Access: roles, guests and grants, private notes, folder limits, agent tokens and their options, share links, invites, signing out everywhere. Anything that lets one person, agent or link reach more than it was given.
- Workspace isolation: reaching another workspace's notes, members or history.
- What runs in the page: the content security policy, plugins and their sandbox, HTML embeds, media embeds, note markdown.
- Secrets: session cookies, agent tokens, Google and bank tokens, share link tokens, and any of them reaching a log, an address, another site or another person.
- Banks: reaching another member's accounts or transactions, or connecting a bank without the passkey step.
- The
inkCLI and the web clipper.
Out of scope
- Preview deployments on
workers.dev. They sign anyone in as a test user on purpose and hold no real notes. A way to turn that on atcommonink.appis in scope. - Denial of service by volume, and rate limits with no further harm.
- Anything that needs the victim's unlocked device, browser profile or Google account, or a malicious browser extension.
- Social engineering, phishing, and physical attacks.
- Google's, Plaid's and Cloudflare's own services. Report those to them. How Common Ink uses them is in scope.
- Reports from automated scanners with no shown impact, such as a missing header or cookie flag on its own, or software versions.
- Self-XSS: script a person can only run against themself.
Testing in good faith
Test against your own account and workspaces, and agents you connected. Don't read, change or delete anyone else's data, don't degrade the service for others, and stop and report as soon as you reach something that isn't yours. We won't pursue work done in good faith within these lines.
How Common Ink protects your data
Common Ink follows a written information security policy, adopted on 6 November 2026 and reviewed every year and after any incident. This section sums it up.
Who can reach it
- Common Ink is run by Audrow Nash LLC. Only its operator, Audrow Nash, has access to Common Ink's servers, its database and its secrets. Nobody else holds a login, key or token to them.
- The policy requires multi-factor authentication on every account that runs Common Ink, from hosting to code to payments. An account signed in to with Google gets it from the Google account's 2-Step Verification, with no password of its own. Any other account has its own, with a passkey or security key wherever the service offers one. None uses text messages.
- It also requires a review of who and what has access every three months, removing anything not needed.
- In the app, a person reaches a workspace only with a role in it or a grant, and an agent only with a token limited to its member's own access or less. Every request is checked against what its person or agent may reach.
Encryption
- Common Ink takes connections over TLS 1.2 or later only, and tells browsers to use HTTPS for it every time (HSTS).
- The database and uploads are encrypted at rest by Cloudflare, which hosts Common Ink.
- Google and bank tokens are encrypted again before they are stored, with AES-256-GCM, under a key the database never holds. Agent, clipper and app tokens are stored only as hashes.
Your banks
- Banks connect through Plaid. Your bank login goes to Plaid, never to Common Ink.
- Before Plaid opens, you confirm it's you with a passkey on your own device, on top of signing in with Google. You manage your passkeys in Settings, under Account.
- Common Ink only reads accounts, balances and transactions. It never moves money.
- A bank is its member's alone. An agent reads it only when you turn on Let agents use accounts, Let agents use transactions or Let agents use envelopes, each for what its page shows.
- Disconnecting a bank deletes what came from it at once and asks Plaid to end the connection. A bank that has needed you to sign in again for 90 days is disconnected the same way.
How changes are made
- Every change to Common Ink goes through a pull request. A new feature is also reviewed by the operator and by a second model or agent that didn't write it.
- Each change runs the typecheck, the unit tests, a dependency audit and browser smoke tests before it merges, and the full browser suite runs on every merge and every night.
- Only the build system deploys, and only from the main branch.
Keeping software up to date
- Dependency updates are proposed every week, and a build fails on any high or critical advisory.
- Once a vulnerability is known to affect Common Ink, it is fixed within 7 days if critical, 30 if high and 90 if medium.
- The code is audited for security at least once a year, and after any large change to access or data handling.
Logs
Logs leave out tokens, cookies, email addresses, share link secrets, sign-in codes, and anything an error quotes. The privacy policy says what is logged and for how long.
If something goes wrong
If we confirm a breach that affects your data, we tell you within 72 hours, and tell Plaid where it affects what came from your bank. Then we fix the cause, add a test that would have caught it, and write up what happened.
Who else handles your data
Cloudflare runs Common Ink, Google signs you in, and Plaid connects banks. Each is added only when needed, after the privacy policy says what it holds, and each is checked every year for a current SOC 2 or ISO 27001 report.
Backups and deletion
The database can be restored to any minute of the last 30 days. So a note purged from Trash, or a revoked token, is gone from the app at once and from the backups 30 days later. An upload purged from Trash is deleted at once, and has no backup. After a restore, every deletion made since the restore point is made again. The privacy policy says what deleting your account removes.
The full policy
Partners who need more for a security review can ask for a reviewer's copy of the information security policy at security@commonink.app.
Changes to this page
This page changes when the policy does. The date at the top says when it last did.