Doc Drift for Confluence

Privacy Policy

Version 1.1 ยท Effective November 5, 2026

This policy explains what data Doc Drift for Confluence (the "App") uses, where it is kept, and how it is removed. The App is made by ARCH BADGER 500, LLC ("we", "us"), an Indiana company, and is sold through the Atlassian Marketplace.

1. The short version

  • The App runs entirely on Atlassian's cloud, on Atlassian Forge.
  • It finds code references on Confluence pages, such as file paths, and flags a page when a merged GitHub pull request changes or removes a file the page mentions. Matching is fixed rules in the App's code. The App does not send your pages to any AI model.
  • GitHub sends the App a notice for each push to your repository's default branch. It uses the push to decide which pages to flag, and keeps only the commit id, the pull request number, the commit time, and the repository, with its owner replaced by a short code (the repository name itself is not changed). It doesn't keep the list of changed files, commit messages, or anyone's name or email address. It never receives or stores source code. When a repository turns on the optional CODEOWNERS input of the Doc Drift GitHub Action, the App may also keep GitHub team slugs (@org/team) for that flag. It still does not receive or store personal names or email addresses.
  • It saves two small content properties on pages it flags, and a third only when an admin turns the freshness marker on (see section 3). It keeps its settings and index in Forge storage inside your Atlassian site.
  • It makes no calls to servers outside Atlassian. We run no servers of our own.
  • It does not store page text, comments, attachments, email addresses, or Atlassian account IDs. It also doesn't store people's names, with two exceptions: a GitHub repository name (owner/repo) can contain a person's GitHub username, and a file path or code name copied from your page can contain a name if your page includes one. A GitHub team slug stored from the optional CODEOWNERS input is not a personal name or an email address, but the slug itself can include a person's name.
  • We cannot read the data the App stores in your site.
  • We do not sell data, and we do not use it for advertising.

2. What the App reads

When a Confluence admin installs the App, Atlassian asks them to approve these permissions:

  • read:page:confluence: read pages, so the App can find the code references on them.
  • read:confluence-content.summary: receive the "page created" and "page updated" events that start indexing.
  • read:confluence-props and write:confluence-props: read and save the App's own content properties (see section 3). Content properties are small pieces of app data attached to a page. This permission does not let the App edit page text, titles, or labels, or add comments. The App's code saves only its own properties and never writes another app's data.
  • search:confluence: search for pages that carry the App's property, using Confluence Query Language (CQL).
  • storage:app: the App's own storage (Forge storage and a Forge SQL database).
  • write:page:confluence: save the App's own content properties through Confluence's v2 page-properties API, and, only for the optional weekly digest an admin turns on per space, create and update the App's own "Doc Drift weekly digest" page in that space. The App's code refuses any other property key, and it updates a page body only when that page is a digest page the App created. It does not edit any other page.
  • read:content-details:confluence: only for the optional weekly digest. The App reads view restrictions on a flagged page and on each of its parent pages, so the digest lists only pages with no view restrictions and shows the others as a count. It reads restriction data only, never changes it, and does not store the restriction lists.
  • read:content.metadata:confluence: only for the optional weekly digest. The App reads a flagged page's parent pages (ancestors) so it can run the restriction check on those parents too. The parent list is not stored.

When a page is created or updated, the App reads the page body and looks for code references: file paths, folder paths, web routes, configuration keys, and environment variable names, plus links to GitHub repositories. It keeps the references and throws the page text away.

The App also receives push notices (webhooks) that GitHub sends to a web address the App provides, once an admin sets this up in GitHub. Each notice must be signed with a secret the admin sets; the App drops notices without a valid signature. A GitHub notice also contains commit messages and the names and email addresses of commit authors. The App reads only the repository name, the branch, the lists of changed, added, and removed files, the first line of the latest commit message (to find the pull request number), the commit id, and the commit time. It does not store commit messages, names, or email addresses.

A repository can send the same kind of signed notice from the Doc Drift GitHub Action instead of, or as well as, a webhook. The Action runs in the customer's workflow. It does not give the App a new permission, and the App still makes no calls to GitHub. If the workflow sets the optional CODEOWNERS input, the Action reads that repository's CODEOWNERS file from the checkout and adds GitHub team slugs (@org/team) for the changed files. The Action drops individual GitHub usernames and email addresses before it sends the notice. If a notice still contains a username or an email address in that field, the App drops it and does not store it.

The App asks for no permission to call servers outside Atlassian.

3. What the App stores

The App stores only these things:

  • Settings. The flag threshold, whether to act only on merged pull requests, a map from Confluence spaces to GitHub repositories, and file patterns to ignore. A Confluence admin types these in. A repository name (owner/repo) includes the name of a GitHub account or organization, which can be a person's username.
  • Webhook secret. The secret the admin sets for GitHub, kept in Forge's secret storage. It is used only to check signatures and is never shown again or logged.
  • Reference index. For each page the App can read: the page ID, the space key, and the code references found on it (type, value, file name, how often it appears, and whether it was in code formatting or plain text). This includes repository names from GitHub links on the page, in plain owner/repo form. For pages the App can't read because of app access rules or page restrictions, it keeps only the page ID, so admins can see how many pages it can't check.
  • Match details. For each flagged page: the repository, with its owner replaced by a short code (the repository name itself is not changed), the pull request number, the score, and the references on the page that matched (file paths and code names exactly as they appear on your page), with the reason for each (for example "file removed"). The App shows these in the page's popup.
  • GitHub team slugs (optional). When the CODEOWNERS input is on, the match details can also include GitHub team slugs (@org/team) for the changed files that flagged the page. The App shows them in the popup. It removes them when someone clears the flag. It does not put them in a content property. A repository webhook that is not the Action does not send them.
  • Feedback. When someone clicks "Not relevant", the references that matched on that page, so they don't flag it again.
  • Processed pushes. The commit id of each push the App handled, with the time and how many pages it flagged, so GitHub's retries are not counted twice. Each record is deleted automatically about 90 days after the push.
  • Page summary (ab500_drift_v1). On each page it flags, the App saves a small summary as a Confluence content property, so you can search it with CQL (for example ab500DriftStatus = "possibly_outdated"). It holds only the status (possibly outdated, verified, or not relevant), a score from 0 to 100, reason codes, the source (github), the repository, with its owner replaced by a short code (the repository name itself is not changed), the pull request numbers, how many flags are open, the relevant times, the matcher version, and the App's name and version. The repository owner is replaced by a short code (the first 16 characters of a SHA-256 hash made with a fixed formula). This is a pseudonym, not anonymization: the same owner gives the same code in every ARCH BADGER 500 app, and anyone who guesses the owner can check the guess. The repository name after the slash is not changed, and it can be searched in Confluence. The summary holds no file paths, page text, titles, names, or account IDs.
  • Byline text (ab500_drift_byline_0ddac1f1). The short text and tooltip shown in the page byline, such as "Possibly outdated since PR #482". When an admin turns the freshness marker on, that tooltip also includes one sentence: "Readers and Rovo should treat this page as possibly stale." The sentence is stored on this byline property. It is not written into the page body.
  • Freshness marker (ab500_drift_freshness_v1). Off unless an admin turns the freshness marker on (markStale). When it is on, the App saves a third content property on a flagged page with only a status (stale or current) and a timestamp. Marking the page "Still accurate" or "Not relevant" sets that status to current only if this App wrote the marker. The marker is not Confluence's own page status. It holds no page text, file paths, titles, or names. Sites that leave the marker off do not get this property.
  • Setup jobs. After install or upgrade, the App puts a job on Atlassian's Forge job queue to set up its database. A job holds only a reason word and no content.

Apart from the cases in section 1, the App does not store page text, comments, attachments, source code, commit messages, or people's names, email addresses, or Atlassian account IDs. Because it stores no Atlassian account IDs, it has nothing to report to Atlassian's personal data reporting service.

4. Where the data is kept

Settings, the webhook secret, match details, feedback, processed pushes, and the reference index are kept in Forge storage and Forge SQL on Atlassian's cloud, in your site's data residency location. Atlassian moves them with your site if an admin changes that location. Setup jobs pass briefly through Atlassian's Forge job queue and hold no content. The content properties are kept on pages in your Confluence site. Atlassian hosts and protects all of it. Atlassian's own privacy policy and Forge data residency documentation apply to that hosting.

Nothing the App stores or reads leaves Atlassian, except as described under AI tools (MCP) in section 5, when your admin and your users choose to connect an outside AI tool. GitHub notices travel from GitHub to Atlassian; GitHub's own terms cover what GitHub sends.

Platform logs. The App writes operational logs to Atlassian's Forge platform. They can include page IDs, space keys, counts, and short error messages (cut to 120 characters), but not page content. The App does not write names or email addresses to these logs itself, but an error message passed on from Confluence, Forge, or GitHub can include an identifier such as a repository name or file path. We can view these logs only to run and fix the App. Atlassian keeps these logs under its own retention policy.

5. Who can see the data

  • People on your Confluence site who can view a page can see its byline flag and open the popup, which shows the PR number, score, and the references on that page that matched. When the flag has GitHub team slugs, the popup shows those too. Anyone who can view the page can click "Still accurate" or "Not relevant".
  • Settings can be seen and changed only by Confluence admins, on the App's settings page. The webhook secret is never shown.
  • Content properties (ab500_drift_v1, the byline text, and the freshness marker when it was saved) can be read by anyone who can view the page, including through CQL search and Confluence's REST API, and by other apps installed on your site that can read content properties, including other ARCH BADGER 500 apps. Team slugs are not in these properties.
  • AI tools (MCP). The App offers two read-only tools through Atlassian's MCP service: one lists flagged pages (at most 50, with each page's ID, title, and space key) and one checks a single page (status, pull request number, score, and the matched references, such as file paths). The tools run as the person using them, so they only return pages that person can view. They never change a page. This is off unless a site admin turns it on. If it is on, the results go to the outside AI tool the person connected, and that tool's own terms apply. The App itself still makes no calls outside Atlassian. Inside Atlassian, custom Rovo agents can use these tools for people who can already see the pages. Sending them to an outside AI tool is off unless a site admin turns it on.
  • We cannot see it. The App sends nothing to us, and Forge does not give us access to the data it stores in your site.
  • App logs. The App's logs record page IDs, space keys, counts, and short error messages. They do not record page text, titles, GitHub notices, names, account IDs, or secrets. An error message passed on from Confluence, Forge, or GitHub can include an identifier such as a repository name or file path. Atlassian controls whether we can see these logs for your site. See Platform logs in section 4.

6. How long data is kept and how it is removed

The App keeps its data while it is installed on your site. Apart from processed-push records, which are deleted about 90 days after each push, it has no fixed retention periods of its own. It replaces a page's references each time the page is indexed, and it updates summaries and match details when a page is flagged or someone clicks "Still accurate" or "Not relevant". Information about a page you delete stays in the App's storage until you uninstall the App. One exception: if a later push matches a deleted page, the App removes that page from its reference index. Its match details and "Not relevant" choices stay until you uninstall the App.

When a Confluence admin uninstalls the App, Atlassian's Forge platform deletes the App's Forge storage and Forge SQL data for your site, and the App can no longer read them. Atlassian's documentation says it keeps a deleted copy for 28 days after an uninstall, so we can ask Atlassian to restore it to a reinstalled App if you ask us within 21 days, and then destroys it under its own retention policy. A reinstall starts with the default settings and an empty index.

The content properties are saved in Confluence itself, not in Forge storage, so they remain on flagged pages after an uninstall, and anyone who can view a page can still read them through Confluence's REST API. A Confluence admin can remove them with Confluence's REST API (DELETE /wiki/rest/api/content/{id}/property/ab500_drift_v1, DELETE /wiki/rest/api/content/{id}/property/ab500_drift_byline_0ddac1f1, and, when the freshness marker was saved, DELETE /wiki/rest/api/content/{id}/property/ab500_drift_freshness_v1). On Confluence's v2 API the same deletes are DELETE /wiki/api/v2/pages/{id}/properties/{property-id} (look up each property's id first). We also provide a script that finds and removes these properties, including after an uninstall (see Removing the App's page properties), or email support@archbadger500.com and we'll help.

While the App is installed, an admin can also find every page the App has marked with the CQL search ab500DriftSchema >= 1. That search stops working after an uninstall; the script can still find the marked pages by checking each page in a space.

Because we cannot access your site's data, uninstalling the App, and removing the content properties as above, is how you remove it. To stop GitHub sending notices, delete the webhook in your GitHub repository settings, and remove the Doc Drift workflow if you added one.

7. When you contact us

If you email support@archbadger500.com, we receive your message and anything you attach. Our support mail is routed by Cloudflare and delivered to a Google mailbox. We use it only to answer you and fix problems. Please don't send secrets, passwords, or real customer data.

8. Our website

Our website, archbadger500.com, uses Cloudflare Web Analytics to count visits. This is separate from the App.

9. Security

The App relies on Atlassian's Forge platform for hosting, storage, encryption, and access control. GitHub notices are checked with an HMAC-SHA256 signature before the App uses them. If we learn of a security incident in the App that affects your data, we will tell affected customers, as set out in our Service Level Agreement.

To report a security issue, email support@archbadger500.com with SECURITY in the subject line.

10. Changes to this policy

If we change this policy, we will post the new version on this page with a new version number and effective date.

11. Contact

ARCH BADGER 500, LLC, support@archbadger500.com

This policy is governed by the laws of the State of Indiana, USA.