Перевод для удобства. При расхождении действует английская версия.
Security
In short: ArtUp Trace is an Atlassian Forge app with the "Runs on Atlassian" designation. There are no ArtUp Labs servers in the path, no outbound network calls, and no third-party trackers. The App reads Jira as the current user, checks Jira permissions on every action, and never writes to your issues. ArtUp Export reads Confluence as the current user, converts pages in the user's browser, stores nothing and never writes to Confluence. ArtUp Reports reads Jira as the current user, builds Excel, Word and PDF files in the user's browser, stores only export templates and never writes to Jira. ArtUp Query reads Jira work items as the app, keeps an index of ids, dates and metadata in Forge SQL and Forge storage, and never writes to work items. ArtUp Release reads Jira as the signed-in user on its pages and as the app for synchronisation and release snapshots, writes only project versions and never writes to issues.
1. Architecture
ArtUp Trace is built entirely on Atlassian Forge and carries Atlassian's "Runs on Atlassian" designation. This means:
- The App's code executes inside Atlassian's infrastructure, not on a server operated by ArtUp Labs. ArtUp Labs runs no external servers for the App.
- The App has no outbound network calls (no egress). It cannot send Jira data anywhere outside Atlassian.
- The App's own data (project settings, requirement and link records, baselines, sync job metadata) is stored only in Forge SQL and Forge storage, in the customer's data residency region as handled by Atlassian.
- The App loads no third-party analytics, trackers or advertising scripts. This website loads no trackers or advertising scripts; it counts page views in aggregate with Cloudflare Web Analytics, which sets no cookies (see the Privacy Policy).
See the Privacy Policy for the full list of what is stored.
1.1 ArtUp Export
ArtUp Export is also built entirely on Atlassian Forge, with no external servers, no Connect modules and only bundled resources.
- Pages, attachments, labels and user names are read in the user's browser, through Atlassian's API and with that user's Confluence permissions. Conversion to Markdown and packing of the zip happen in the browser, and the zip is downloaded to the user's device.
- The app has no outbound network calls (no egress). Confluence content is not sent to the app's backend, to ArtUp Labs or anywhere outside Atlassian and the user's browser.
- The app stores no data: no Forge storage is used, although the
storage:appscope is declared for future versions. - Its only backend function returns whether the licence is active. It receives no page content, so its Forge logs contain no Confluence content.
1.2 ArtUp Reports
ArtUp Reports is also built entirely on Atlassian Forge, with no external servers, no Connect modules and only bundled resources (fonts for Latin, Cyrillic and CJK text, libraries and images).
- Issues, attachments and user names are read in the user's browser, through Atlassian's API and with that user's Jira permissions. The Excel, Word or PDF file is built in the browser and downloaded to the user's device.
- The app has no outbound network calls (no egress). Issue content is not sent to the app's backend, to ArtUp Labs or anywhere outside Atlassian and the user's browser.
- The app stores only export templates in Forge storage: names, settings, Excel column sets, uploaded .docx template files and the Atlassian account id of each template's author. No issue content is stored or logged.
- Its backend functions manage templates and check the current user's Jira permissions. On failure, Forge logs contain only the function name and the error message.
1.3 ArtUp Query
ArtUp Query is also built entirely on Atlassian Forge, with no external servers, no Connect modules and only bundled resources. It adds JQL functions to Jira (jira:jqlFunction), a reference and status page, and an administration page.
- The app has no outbound network calls (no egress). Work item data is read through Atlassian's API and never leaves Atlassian.
- It reads work items, links, the hierarchy, the changelog for Sprint and status, and comment and attachment metadata as the app, and never writes to work items, comments or issue properties. The results of a function are the same for every user and are applied by Jira to the user's own search, so Jira hides the work items the user cannot see in the search result. The next point says what that does not cover.
- A subquery runs with the app's access, not the user's. A function is computed once for the whole site, as the app, and Jira applies the result to each user's search. A work item the user cannot see is never returned: Jira removes it from the result. But the result can still reflect it. The functions that follow links or the hierarchy (
linkedIssuesOfand its recursive forms,subtasksOf,parentsOf,epicsOf,issuesInEpics,childIssuesOf) read their subquery across all projects, including projects the user cannot browse. A work item the user can see is returned when it is linked to, or is the parent, child or epic of, a work item in such a project that matches the subquery. A user who can write JQL can vary the condition and learn from that whether a hidden work item that matches it exists, and the message for an invalid subquery can show whether a project exists. Excluding a project on the administration page removes its work items from results; it does not stop them from being the source of a subquery, so it does not close this. The summary, key and content of a hidden work item are not disclosed. - The app stores in Forge SQL and Forge storage: ids, dates, status categories, sprint names, the account ids of comment and attachment authors, the visibility of comments (type and the name or id of the role or group), file extensions, cached result ids and, while a function is in use, its arguments — the JQL of subqueries, a user name or email address given to
by, dates and expressions. The arguments are held in the update queue (about ten minutes after the last call), in the queue of large results (until computed, at most six hours) and in a cached list of function calls that is rewritten about hourly. The app keeps no copy of the hierarchy or of the links. Issue text, comment bodies and attachment content are never stored. Comments with restricted visibility are ignored by every function. - Only Jira administrators can change the settings (excluded projects, reindex, rebuild). The app's error log stores the function name and a fixed message without values, never the function arguments.
1.4 ArtUp Release
ArtUp Release is also built entirely on Atlassian Forge, with no external servers, no web triggers, no Connect modules and only bundled resources. It adds a Jira page (jira:globalPage), an administration page (jira:adminPage), triggers for version and project events, background queues and an hourly scheduled check.
- The app has no outbound network calls (no egress). Version and issue data is read through Atlassian's API and never leaves Atlassian. Word and PDF release notes are built in the user's browser from data the page already shows.
- The pages read and act as the signed-in user: a user sees only the programs whose lead project they can browse, and projects and issues they cannot see are shown as “No access” or not counted. Creating, editing, releasing and archiving versions from the page writes to the lead project as that user and is offered only with the Administer projects permission there; deleting copies of a deleted version needs Administer projects in that project; confirming a released or archived change in a copy needs a Jira administrator. Programs are managed on the administration page, which only Jira administrators can open.
- Synchronisation, the sync checks and release snapshots run in the background as the app. A snapshot is taken as the app so that it is complete whoever pressed Release, and it is shown to each user filtered by what that user can see now: hidden rows are counted, not shown. A project the app cannot read makes the snapshot “Incomplete” and is named.
- It writes only project versions: it has no scope to change issues and does not change them. The sync never deletes a version; deleting the copies of a deleted lead version is a user's action, and the user chooses where their issues go.
- The app stores in Forge SQL and Forge storage: program definitions (name and project ids), links between versions with the last synced name, description, dates and released and archived flags, a sync log of the newest 1,000 entries per program (no issue text), release snapshots (issue id and key, project id, issue type, summary and status category), matching previews for 24 hours and working state. No issue descriptions, comments, attachments, custom field values, user names, email addresses, avatars or Atlassian account ids.
- Forge function logs contain only fixed messages and counters — the failed step and the error class or HTTP status, hourly counts of Jira requests, and program ids when a check is skipped. No issue content, version or project names and no user data are written to the logs.
2. Access scopes and permissions
The App follows the principle of least privilege. It requests only the Atlassian scopes it needs, and nothing more:
| Scope | Used for |
|---|---|
read:jira-work | Read issues, issue links and project data needed to compute coverage and detect suspect links. |
read:jira-user | Read basic information about the current user, used for the per-action Jira permission check. |
storage:app | Store the App's own settings, requirement/link records and baselines in Forge storage. |
ArtUp Trace requests no write scopes for Jira issues. It reads Jira data as the current user, so it can only see what that user is already permitted to see in your Jira site, and it never writes to Jira issues.
Beyond the scopes above, the App performs a Jira permission check on every user-facing action before it acts — for example before showing an issue's traceability data or before recording a link confirmation — so a user cannot use the App to see or affect anything their own Jira permissions would not already allow.
ArtUp Export requests only read scopes. It reads Confluence as the current user, so pages the user cannot see are not exported, and it never writes to Confluence.
| Scope | Used for |
|---|---|
read:page:confluence | Read pages, their content and versions. |
read:space:confluence | Find the space and its top-level pages. |
read:hierarchical-content:confluence | Read the page tree in Confluence order. |
read:attachment:confluence | List and download page attachments. |
read:label:confluence | Read page labels for the front-matter. |
read:confluence-user | Read display names of page authors and mentioned users. |
search:confluence | Find pages by title in the page picker. |
storage:app | Declared for future versions; not used by the current version, which stores nothing. |
ArtUp Reports requests only read scopes and storage:app. It reads Jira as the current user, so issues the user cannot see in Jira are not exported, and it never writes to Jira. Before a user creates, changes or deletes a project or site template, the backend checks that user's Jira permissions (project administration for a project template, Jira administration for a site template); a personal template is visible only to its author.
| Scope | Used for |
|---|---|
read:jira-work | Read issues (search, fields, comments, work logs, links), saved filters, attachments and their thumbnails. |
read:jira-user | Read display names of authors, assignees and template authors. |
read:board-scope:jira-software | Read the issues of a board or backlog. |
read:sprint:jira-software | Read the issues and details of a sprint. |
read:board-scope.admin:jira-software | Read board configuration, to find the filter behind a board. |
read:project:jira | Read the project list for the templates tab and project templates. |
storage:app | Store export templates in Forge storage. |
ArtUp Query reads Jira as the app and requests one write scope, write:app-data:jira, which is used only to store the app's JQL function precomputations. It requests no write:jira-work and no manage:* scope, and it never writes to work items.
| Scope | Used for |
|---|---|
read:jira-work | Read work items, links, the hierarchy, the changelog (Sprint, status), comment and attachment metadata, and run JQL searches. |
read:jira-user | Resolve the users that conditions such as by and inRole mean. |
read:user:jira | Read group members for inGroup (together with read:group:jira). |
read:group:jira | Read groups and their members for inGroup. |
read:avatar:jira | Comes with the user and group reads; no avatars are stored. |
read:board-scope:jira-software | Find boards for the sprint functions. |
read:sprint:jira-software | Read sprints (ids, names, states). |
read:project:jira | Read the project list for the settings page and to check project keys. |
read:app-data:jira | Read the state of the app's JQL function precomputations. |
write:app-data:jira | Store the app's JQL function precomputations. Not a write to work items. |
storage:app | The app's own Forge storage: the update queue, settings and the error log. |
ArtUp Release requests no write:jira-work and no read:jira-user. Its one write scope for Jira, manage:jira-project, is used only for project versions.
| Scope | Used for |
|---|---|
read:jira-work | Read project versions and their details, the project list, issue counts and the issues of a version (by version id), and the permissions of the app and of the user in a project. |
manage:jira-project | Create, update, release, archive and delete project versions in member projects, and in the lead project when a user acts from the page. Not a write to issues. |
storage:app | The app's own Forge storage (working state, matching previews). Forge SQL needs no scope. |
3. Application security controls
- Permission checks on every action. Every user-facing operation re-checks the calling user's Jira permissions at the time of the request; nothing is cached in a way that could grant stale access.
- Input validation. All input from the front end, from Jira webhooks, and from imported data (for example CSV or configuration values) is validated before it is stored or used to build a query, to guard against malformed or malicious input.
- No egress, no trackers. Because the App cannot make outbound network calls, there is no path for it to exfiltrate data even if a dependency were compromised.
- ArtUp Export: untrusted files are checked. A previous export dropped in for an update is read in the browser, and paths from its manifest that could point outside the export folder are refused.
- ArtUp Reports: uploaded templates are checked. A Word template must be a valid .docx file of at most 2 MB that unpacks to no more than 20 MB, and its placeholders are checked before it can be saved.
- ArtUp Query: arguments are not logged. Function arguments contain the customer's JQL, so the error log keeps only the function name and a fixed message without values; the arguments themselves are kept only while the function is in use (section 1.3). Conditions, dates and expressions are parsed by the app and refused with a message when invalid.
- ArtUp Release: state, not events. The app compares the current state of the lead version with the last state it applied and writes only the difference, with at most one write at a time in each project, so that duplicate or out-of-order Jira events do not create duplicate versions. A name already taken in a member project is reported, not overwritten.
- Least-privilege storage. Only the fields needed for traceability are stored; see the Privacy Policy for the exact list. Full issue description text is never stored, only a change-detection hash.
4. Dependency and supply-chain security
- Automated scanning.
npm auditruns as part of the build, and Dependabot is enabled on the source repository to flag known vulnerabilities in dependencies and open update pull requests automatically. - Review before release. Dependencies are checked for known vulnerabilities before every release, not only on a schedule.
- Minimal dependency footprint. The App uses as few third-party packages as practical, to reduce supply-chain exposure.
5. Account and workstation security
- Multi-factor authentication (MFA) is enabled on the developer's Atlassian, source control, email, DNS and app-store accounts.
- Full-disk encryption and automatic OS updates are enabled on the development workstation.
- Unique, generated passwords are used and kept in a password manager.
- Secrets and credentials are kept out of source control and rotated periodically and immediately after any suspected exposure.
6. Reporting a vulnerability
If you believe you have found a security vulnerability in ArtUp Trace, ArtUp Export, ArtUp Reports, ArtUp Query, ArtUp Release or on this website, please email [email protected] with:
- a description of the issue and its potential impact;
- steps to reproduce it, including any request or response details that help us confirm it;
- whether any customer data was, to your knowledge, affected.
Please do not include Jira or Confluence content beyond what is strictly needed to demonstrate the issue, and do not test against customer sites you do not control. We will acknowledge your report within 24 hours and keep you updated as we investigate and fix it. We currently do not run a paid bug bounty programme.
7. Incident response
ArtUp Labs follows a written incident response plan for every Marketplace app, owned by the security contact below and reviewed at least once a year and after every incident.
7.1 What counts as an incident
- Unauthorised access to customer data held by an app (Forge SQL, Forge storage).
- A vulnerability in an app that exposes customer data or lets a user act beyond their Jira or Confluence permissions.
- Compromise of a developer account, workstation, source repository or deployment credentials (Atlassian account, Forge token, source control, domain/DNS, email).
- A report from Atlassian (Marketplace Security ticket, bug report) or from a customer about any of the above.
7.2 Detection channels
- [email protected] and [email protected], checked daily on business days.
- Atlassian Marketplace Security (AMS) tickets on ecosystem.atlassian.net.
- The Forge developer console: app logs, invocation errors and alerts.
- Dependency alerts (
npm audit, Dependabot).
7.3 Response steps
| Step | Target time | Action |
|---|---|---|
| Acknowledge | 24 hours | Confirm receipt to the reporter; open an internal incident record (date, source, affected apps, versions). |
| Triage | 24 hours | Rate severity (Critical / High / Medium / Low); decide whether customer data is affected. |
| Notify Atlassian | within 24 hours of becoming aware of an incident affecting customers | Raise a P1 ticket with Atlassian Marketplace Security and keep it updated until closed. |
| Contain | as soon as possible | Rotate compromised credentials (Atlassian API tokens, Forge credentials, source control, DNS, email); revoke sessions; if needed, ship a version that disables the affected feature, or ask Atlassian to pause the app. |
| Fix | within the Marketplace Security Bug Fix Policy due dates for the severity | Patch, test, deploy to production, and confirm the fix with the reporter or Atlassian. |
| Notify customers | within 72 hours of identification, when their data is affected | Email the technical and billing contacts of affected installations: what happened, what data, what we did, and what they should do. |
| Close | after the fix is verified | Write a short post-incident review: root cause, timeline, and what changes prevent a repeat. |
7.4 Preventive controls
- Full-disk encryption and automatic OS updates on the development workstation.
- Multi-factor authentication on Atlassian, source control, email, DNS and app-store accounts.
- Unique passwords kept in a password manager; secrets kept out of source control and rotated regularly and after any suspected exposure.
- The Apps run on Atlassian (Forge) with no external egress, least-privilege scopes, and a permission check on every user-facing operation.
- Dependencies are checked for known vulnerabilities before every release.
8. Contact
Security contact: [email protected]. For general questions, see our Support page. For what data is stored and how, see our Privacy Policy.