What we need from you:

• The issue: What is happening vs. what you thought would happen.
• The location: A link to the affected forums page or thread.
• The visuals: Screenshots and/or error messages. (These really help ensure we are looking at the exact same thing).
• Your setup: Your browser and Operating System, and whether you are on mobile or desktop.

👉 Important: We can’t fix issues inside Atlassian products like Jira, Confluence, Trello, etc.
If you’ve found a bug in an Atlassian product, please visit our public bug tracker to see if it’s already reported. You can watch or vote on any matching issue. This thread is just for Community forums bugs.

The good news: Because our new platform finally supports threaded replies in Articles, the Community team and moderators can now reply directly to your comment! We will chime in to ask for more details, confirm we have recreated the bug, or celebrate with you when it is fixed.

We are still unpacking a few boxes around here, and your eagle eyes make a huge difference. Thanks for pointing out the bumps in the road so we can smooth them out for everyone else!

How can it possibly be this difficult to allow people to USE YOUR SOFTWARE.

Hi Atlassian Support,

I’m unable to see my previous documents in Confluence. The documents were available earlier, but I can no longer find or access them.

Could you please help me check:

• Whether the previous Confluence pages/documents still exist
• If they were accidentally deleted or moved
• Whether there is any permission or access issue
• How can I recover or restore the previous documents

Please let me know if you need any additional details from my side.

Summary of the Issue

There is a noticeable race condition between human reaction/click execution time and asynchronous state transitions within the shared user-picker dropdown component.

When clicking into an assignee or user-mention field, the dropdown immediately renders local static options (Assign to me, Unassigned, Automatic). In everyday usage, this menu often remains static just long enough (~300ms–600ms) for a user to read the option, commit to an action, and initiate a physical mouse click.

However, once the background directory query resolves, the component destroys the initial layout and re-renders the global directory results right as the mouse click lands. Because click events register sequentially, the click coordinate hits an injected directory user rather than the intended target, leading to accidental reassignments and permissions changes.

Steps to Reproduce

1. Focus an assignee field, page restriction modal, or inline task user-mention box.
2. Notice the initial static placeholders (Assign to me / Unassigned) appear immediately.
3. Pause for ~300ms–500ms (standard cognitive read-and-click time).
4. Move the cursor and click Assign to me.
5. Observe the background directory query payload resolve and re-render the dropdown contents right as the click registers.

Actual vs. Expected Behavior

• Actual Behavior: The layout shuffles underneath the mouse pointer mid-click, causing the user to accidentally select a directory user injected into those exact pixel coordinates a split-second prior.
• Expected Behavior: The visual positions of default actions should remain stable and uninterrupted while interacting with the menu.

Suggested Technical Mitigations

1. Static Placeholder Locking: Lock early-rendering default elements (Assign to me / Unassigned) into a permanent, immutable top row so late-arriving asynchronous results append strictly below them.
2. Click-Blanking Safety Buffer: Introduce a brief input-blocking buffer (100ms–150ms) following an asynchronous DOM layout shift to discard unintentional mis-clicks.
3. Hover-Freeze State: If the cursor is actively hovering inside the dropdown bounding box, delay rendering incoming asynchronous search results until the cursor becomes stationary or exits.
   

Has anyone else encountered this race condition or found workarounds? Hoping the Atlassian frontend / design system team can review this component behavior.

The only way I know of to not show "experimental" or "learning how to build am Agent" is to set the Open to all users to off. Has any one else found this as open door to floor the Agent lists, and if so, how have you dealt with it?

If the Open to all users was by default off, users could experiment until their hearts content and then use an inhouse governance process to get approve to turn the option on. Not having the Agent lists in alphabetical order doesn't help either, but that is for another day. I appreciate you comments, suggestions, and input.

For those of you who deal with renewals: when you're looking at an app you
suspect nobody uses, what would you actually need to see before you cancelled
it with confidence?

Specifically curious about:

• Would you trust "nobody has opened it in 90 days" more, or "nothing in our
  configuration references it"? Or neither on its own?
• When an app has left custom fields, post-functions or dashboard gadgets
  behind, how do you find all of that today?
• Has a scream test ever gone badly for you?
• Do you get a real say at renewal, or does the invoice just get paid?

To be upfront: I'm a developer, not a vendor. I have nothing built and nothing
to sell — I'm trying to work out whether this is a genuine day-to-day problem
or just an annoyance people have learned to live with. Happy to share back
what I learn.

I am trying to automate a Rovo agent in Jira Cloud (team-managed project). Requirement:

When a work item of type Story is moved to status In Progress, the agent should generate test cases and those tests should be created automatically in Jira (as Subtasks on that Story). I should not have to copy JSON from Chat or create the tests by hand.

What I have done

1. Created a custom Rovo agent in Studio (instructions: MODE CREATE_TEST_CASES, return JSON only with a newTests array: summary, priority, type, description).
2. From the agent, I used “Add automation trigger”.
   • Trigger text: When a work item in my project is transitioned to In Progress and the issue type is Story.
   • Instructions: MODE CREATE_TEST_CASES, return JSON for {{issue.key}} so automation can create subtasks.
3. On the rule that opened, I added the “Create sub-tasks” action:
   • Work item: Current work item
   • Type: Subtask
   • Summary: I am not sure what to put here ({{test.summary}} vs a fixed title).

What is happening

• Moving the Story to In Progress sometimes runs the agent and I get JSON in Chat or in agentResponse.
• Subtasks are not created (or only a blank/wrong subtask is created).
• I understand from docs that if an agent is used in an automation flow it cannot use its own tools and only returns text via {{agentResponse}}.
  https://support.atlassian.com/studio/docs/add-tools-to-rovo-agents/
  https://support.atlassian.com/rovo/docs/agents-in-automations/

Questions

1. For this use case, should I use the Studio “Add automation trigger” wizard, or a normal Jira Automation rule with the “Use Rovo agent” action?
2. After “Use agent”, how do I create one Subtask per item in the JSON array? Is the correct pattern:
   For each / Advanced branching on {{agentResponse.asObject.newTests}}
   then Create work item / Create sub-tasks with {{test.summary}} and {{test.description}}?
3. Does “Create sub-tasks” (current work item) support looping over agentResponse, or do I have to use Create work item inside a For each branch?
4. If the agent returns markdown around the JSON, will {{agentResponse.asObject.newTests}} be empty? Any tips to force JSON-only output?
5. Team-managed project has no Test issue type — is Subtask on the Story the recommended type for auto-created tests?

Project type: Jira Software Cloud, team-managed. Agent is custom (Studio). I am not asking for an exploit or anything outside my own site.

Thanks.
Subash B

We run a code review in Bitbucket Agentic Pipelines with provider: claude. The agent step can't use current Claude models. As far as we can tell the cause is the Claude Code version the installer bundles, not our configuration.

What we see

• claude-sonnet-5 fails with a 400 that asks for thinking.type: "adaptive".
• claude-opus-5-5 fails the same way. The client sends thinking.type: "enabled", and current models reject that at every effort level.
• Get past that, and the API refuses the client itself: "version 2.1.280 or newer is required".

Where it comes from

The Claude step runs bitbucket-pipelines-agent-installer:7-claude. The ACP handshake in the step log reports claude-agent-acp 0.25.3. On npm, that version pins @anthropic-ai /claude-agent-sdk 0.2.91, which carries Claude Code 2.1.91. The old client picks thinking.type: "enabled" from its own model table, and it is below the 2.1.280 minimum that current models need.

The current release, claude-agent-acp 0.81.2, pins claude-agent-sdk 0.3.280, which carries Claude Code 2.1.280.

What we tried

• Installing @agentclientprotocol /claude-agent-acp@0 .81.1 in the step before the agent: line. The handshake still reported 0.25.3, so the installer's own copy is what runs.
• A proxy that rewrote thinking.type: "enabled" to "adaptive". That got past the 400, and the next request was refused on the Claude Code version.

The same pipeline with provider: codex runs current OpenAI models without problems.

Questions

1. Is there a timeline for updating the Claude installer to a current claude-agent-acp?
2. Is there a supported way to choose the installer version, or to run our own claude-agent-acp, in a pipeline?
3. Could the installer release notes list which claude-agent-acp and Claude Code version each tag bundles? Then we could tell from the tag when current models become usable.

Noteworthy to mention that we upgraded to Jira v11.3.8 a few months ago, and these complaints started appearing after the upgrade. However, we cannot confirm whether the issue is related to upgrade.

Also, note that we heavily use Script Runner behavior to implement custom need on the screens. However, there is no logic in the Behavior scripts that is to wipe out fields or clearing fields.

So, wanted to reach out here to see if anyone else in the community experienced similar behavior? If yes, what was the work around that you took to remedy this annoying issue?

~ Alauddin Azad

I need help recovering access to my original Trello/Atlassian account.

The email address associated with the account is:

wildnfreeWHS@gmail.com

I am the owner of this account and I still know the correct password. I can successfully enter the email address and password, but Atlassian then requires me to enter a six-digit identity verification code that is being sent to the old email address.

Unfortunately, the Gmail account wildnfreeWHS@gmail.com was permanently deleted while I was out of the country, and I no longer have any way to access that mailbox or receive the verification code.

I have already created a separate Atlassian/Trello account using my new email address, but I need to regain access to my original account because my original Trello boards and information are associated with that account.

I am not asking to create a new Trello account. I need help recovering my original Atlassian account and, if possible, changing the account email address to my current accessible email address.

I can provide additional information to verify ownership of the account if needed.

this account was a was not associated with any company.

Any help would be greatly appreciated.

Thank you.

G'day, community!

We're writing to let you know about upcoming changes to dashboard subscriptions in Atlassian Analytics, Assets Reports, and JSM Operation Reports.

Starting September 8th, 2026, you will no longer be able to create new dashboard subscriptions. On November 2nd, 2026, all existing subscriptions will stop being sent, and the subscription feature will be fully retired.

What's NOT changing

• Chart CSV/PDF export - exporting chart data as CSV/PDF continues to be supported.

• Dashboard CSV/PDF/JSON export - exporting dashboard data will still be available.

What you should do

I need your help to check if it is possible to validade a request type value in an IF clause in an automation rule. For example:

The Request type called Observabilidade needed to be validated in order to set set another field values and create other linked issues.

It doesn't work neither in the IF clause nor in the ELSIF clause.

Could you plase help me on this?

Thank you.

Site connection info

Note: To participate in this workshop, you will need to use the dedicated workshop site and login credentials provided on your handout (not your daily work account).

1. Open a browser window in Incognito / Private mode (or use a clean browser profile).
2. Navigate to the site URL listed on your login sheet.
3. Sign in using the assigned username and password.

Learning Session Overview

🖐️ Welcome to this learning session. During this workshop, we’ll build a custom Jira dashboard using Studio from start to finish.

This is just one example of what Forge makes possible. Imagine what you could build next.

What you'll build

A Jira dashboard tailored to a specific space, featuring:

• A 2D stacked bar chart showing team workload distribution
• A panel of unassigned work items by priority for quick triage

What you'll learn

By the end of this workshop, you will be able to:

• Define clear app specs for your dashboard, using annotations and an image to show Rovo what you want.
• Build a custom Jira dashboard using Forge, with data visualizations powered by live space data.
• Iterate on your app by refining your specs and dashboard based on what you learn.
• Publish the app to a test site, with the ability to restore a previous version as needed.
• Replicate the end-to-end flow from idea → Forge app → live dashboard.

First-time access to the site

When prompted during onboarding, click on Other on the screen below and you are all set.

AI is non-deterministic by nature

Let’s start: app specifications

1. Select the app switcher in the top left corner (the one with the four little squares ).
2. Press Rovo Studio ( ). If you don’t see it, click the … more menu first, then select Rovo Studio.

Define the app specs

1. In the prompt text area, enter the following prompt and continue by either pressing Enter/Return or clicking the upward-pointing arrow:

I need a Jira dashboard (Jira project page) showing key information for a specific space. The dashboard is available on every space.

Include a 2D stacked vertical bar chart showing team workload distribution by assignee, color-coded by work item status, and a table showing unassigned work items by priority.

While defining the specs, Studio may ask clarifying questions. If prompted, reply by either entering the number of your choice or typing your response in the Tell Rovo what to do prompt box.

Expand to see a few examples

Answer: 1, Single project per dashboard

Answer: 1, View-only dashboard

Review the app specs

1. Click on the Review app specs button to build and visualize the specs as currently defined.

Build the app

1. Press Build app for Rovo Studio to create the app and generate the preview.
   1. If the Build app button is not available (e.g. because a subsequent prompt has been added), simply type Build app in the text area.
2. Click on Select developer space → Select an existing space → pick Team 26 from the drop down → click on Assign to developer space.

Let’s iterate on the app: annotations

1. Click on the Annotations button ( ) located in the top-right toolbar of the preview area:
2. Select the topmost heading in the UI (e.g., Team workload or Workload).
3. Enter the following text and continue by either pressing Enter/Return or clicking the or clicking the upward-pointing arrow:

Change the label to “Hello World from Amsterdam”

4. In the chat box, either press Enter/Return or click the upward-pointing arrow to submit the annotation request prompt.

Revert the change

1. In the chat box, hover over the App Built panel and click on the Restore to draft arrow () to restore to a previous checkpoint

2. In the Restore checkpoint? dialog, click Restore.

3. Once done, click on the three dots (…) in the top right corner and select Version history to display the versions of this app.

Let’s iterate on the app: image

1. Download the image below by right-clicking on it and using the Save Image As… on your browser.

2. Drag & drop the locally available image into the chat box.
3. In the chat box, enter the following text:

Reorder the columns to match the picture

4. Either press Enter/Return or click the upward-pointing arrow to submit the image change request.

Publish the app

1. Press Publish app for Studio to install the app on the test site
   1. If the Publish app button is not available (e.g. because a subsequent prompt has been added), simply click on the Publish option in the top right area of the screen
2. Select your test site in the Publish to production dropdown → click Publish (do not change any of the other options)

View the app in Jira

1. In a new browser tab, go to:
   https://<your-site-subdomain>.atlassian.net/browse/SALES
2. Replace <your-site-subdomain> with the site name assigned to you on your workshop credential sheet (for example: one-atlas-abc.atlassian.net/browse/SALES).
3. Find the Team Dashboard board that was created and installed, and click it to open. The dashboard might appear under the More menu.

👏 Great job in defining, creating, iterating and publishing a Forge app!

Caterina

https://www.linkedin.com/in/caterinacurti/

https://x.com/caterinacurti

On the Backlog Insights > Sprint Commitment > Total Complete, I've found a couple of examples where it is over 100% because the team completed more points than they committed to at the start of the sprint. However, I've yet to find an instance across any of my teams where it was less than 100%, even when I know that they completed less than they committed to. In some cases, the stories were moved out before closing the sprint. In other cases, the stories were moved as part of closing the sprint. In both cases, the % complete should be less than 100%, but it never is. How is that being calculated?

Hi,

Is it possible to get all object with an attribitute superior at 1 data?

For exemple, I want all group with more than one user associeted (user one attribut of group object).

Thank you in advance for your help.

Regards,

Mickael O. 

All our projects are created using a standardised project template with a common set of workflows and fields. However, some projects do not require most of these fields.

In our on-premises environment, we used the CreateIssueDetails URL with query parameters to pre-populate fields. This allowed us to provide a customised creation form containing only the fields relevant to each application. However, in Jira Cloud, the CreateIssueDetails screen displays all fields configured for the Create screen instead of showing only the required fields and preset values.

We have found an alternative using space forms, which allows us to create customised forms by dragging and dropping the required fields. However, space forms do not currently support preset default values, making the process less streamlined for our users.

We understand that URL-based field pre-population and post-creation automation may serve as alternative solutions. However, we do not consider these to be sustainable or user-friendly approaches in the long term.

Could the Atlassian team please provide an update on whether support for preset default values in space forms is planned and, if so, when it is expected to become available?

Hi everyone,

We're on Jira Cloud Free (cinewav-sj.atlassian.net) and hit the 2 GB storage limit. We deleted a large number of attachments more than 72 hours ago, but storage usage hasn't decreased and we still can't upload new files.

When I open admin.atlassian.com → Apps → Sites → Storage, the page shows "Something went wrong. Try again later." (screenshot attached). The page itself says changes should appear in Jira within 30 minutes.

Could someone from Atlassian please check and recalculate storage for our site?

Thanks!

Hi everyone,

I’m using Trello to manage local SEO work for service-based businesses, and I’m trying to improve how I handle recurring SEO checks.

For example, when managing a local service website, there are things I don’t want to check only once. I may need to review:

• Google Business Profile optimization
• NAP consistency
• important service pages
• broken or missing internal links
• title and meta changes
• local citations
• new reviews and review responses
• indexing of newly published pages
• backlink status
• technical issues that appear over time

Right now, I create cards for these tasks, but after marking them complete, I sometimes lose visibility on when they need to be checked again.

I’m thinking about using Trello automation, due dates, recurring cards, or custom fields, but I’m not sure what would be the cleanest setup without turning the board into a collection of repetitive cards.

For those managing ongoing marketing or website maintenance projects in Trello:

How do you handle recurring tasks that need to be reviewed weekly, monthly, or quarterly?

Do you use Butler automation, recurring due dates, templates, or a separate maintenance board?

I’d especially appreciate a setup that works well when managing several service-business websites at the same time.

Thanks!

How can Confluence Version History support a formal review and sign-off process?

We use a live Confluence document to manage controlled documentation. Changes are made directly in the document, and the rationale for those changes is discussed through comments. A manager reviews the comments, provides feedback, and either accepts or rejects the proposed changes.

We also maintain a changelog within the live document using a database. After all changes have been reviewed, the document must go through a formal sign-off process.

In the past, we managed documents in Microsoft Word using Track Changes. This made it very easy for the sign-off manager to see exactly what had changed between approved versions. We are trying to achieve a similar level of transparency in Confluence, but I have a few concerns regarding Version History:

1. Creation of versions seems unpredictable

Confluence appears to decide automatically when a new version is created. Sometimes a new version is generated when multiple people are working in the document, while at other times similar edits do not seem to create a new version immediately.

In addition, there can be a considerable delay before a new version appears.

This creates challenges for our review process because:

• Multiple authors may be working in the document simultaneously.
• A single review cycle can result in dozens of versions.
• It becomes difficult for the sign-off manager to identify which versions should actually be compared.

Is there a way to manually create, label, or request a new version at a specific point in time, similar to creating a controlled revision?

2. Compare Versions behaves inconsistently

When using Compare Versions, the output differs between documents.

In some live documents, almost all content appears as struck through, which makes it look as though everything has been removed. In other documents, changes are displayed much more clearly, with additions highlighted in green and text shown as inserted, modified, or deleted.

As a result:

• I am not always confident that I am interpreting the comparison correctly.
• I am unsure whether all changes are actually being displayed.
• I do not know how to verify that the comparison is complete and accurate.

Is this a known limitation of Confluence's comparison engine, particularly for Live Docs or pages edited by multiple users?

3. Best practice for formal approvals

For organizations that require document review, approval, and sign-off:

• Do you rely on Version History alone?
• Do you maintain a separate changelog?
• Are there recommended apps or workflows that provide functionality similar to Microsoft Word's Track Changes?

I would appreciate any guidance or examples of how others manage controlled documents and sign-off processes in Confluence.

• API response: size: [13], isLastPage: true. There is no next page.
• Project settings → Customers page in the UI: [50] customers.
• Difference: [37] customers appear in the UI but not in the API response.

The 13 returned customers are the first 13 rows of the UI list. Every customer after that is missing, including customers with a visible email address.We also tried the same call as a logged-in user and via a Forge app, with the same result.

Expected behaviour

The API should return every customer shown on the project’s Customers page, or the documentation should explain which customers are excluded and why.

I've been really excited about the new Individual Capacity Planning View as it's something that's been a missing cog for a long while.

But it's still very very basic - really keen to understand if there is a development roadmap and what that looks like and when?

Also very happy to contribute feedback

I'm trying to model dependencies across different product releases in Jira so we can create a master roadmap and clearly visualise delivery critical paths.

The challenge is that a release (<span>fixVersion</span>) cannot directly have issue links such as <span>blocks</span> / <span>is blocked by</span>. Because of this, I'm considering introducing another issue type that acts as a container for all releasable features planned for a given release.

For example, something like a "Release Package" issue type with:

• Assignee
• Start date
• Due date
• Description
• Ability to use issue links (<span>blocks</span> / <span>is blocked by</span>)
• Ability to associate a release/version
• Ability to associate one or more products/components

The proposed hierarchy would look like this:

• Release Package
  • Epic / Releaseable Feature
    • Story / Bug / Task

The main goal is to:

• Track dependencies between releases
• Visualise cross-team delivery risks and sequencing
• Build a higher-level roadmap view beyond standard fixVersions

Has anyone implemented something similar in Jira?
Would you recommend:

• creating a custom issue type like this,
• using Advanced Roadmaps differently,
• or modelling releases/dependencies another way?

Site: mydigitaloffice.atlassian.net
Project: TPS (service desk ID 3115)
Channel: Embeddable widget, widget type, widget version 2026-09-28_12-27

Every message returned by the virtual service agent in the embeddable widget renders as a card reading "Unsupported doc" instead of the answer text. This affects all bot replies—knowledge-base answers, fallback messages ("Hmm, I'm still having trouble understanding"), and email-capture prompts alike.

The same agent answering the same questions on the customer portal renders perfectly.

Critically, this reproduces inside Atlassian's own admin UI, with no customer code on the page:
Project settings → Channels → Widget → the live preview panel. That rules out our embed script, our page, and our CSP.

Steps to reproduce

1. Open Project settings → Channels → Widget for a JSM project with the virtual service agent enabled.
2. In the built-in preview, send any message (e.g., "What are all the plans in TruePlan?").
3. The agent responds—Yes/No feedback buttons appear, so generation succeeded—but every message block renders as a card reading "Unsupported doc."

Expected vs. actual

• Expected: the agent's reply text renders in the widget, as it does on the portal.
• Actual: one "Unsupported doc" card per message block. No answer text is visible to the end user.

Our analysis (payload + bundle inspection)

We captured the widget's network traffic and read the published widget bundle. The server response is correct—the failure is client-side, in the widget's rendering path.

1. The server sends valid ADF. Response from authorType: "bot", contentType: "ADF":

{"type":"doc","version":1,"content":[
  {"type":"paragraph","content":[
    {"type":"text","text":"No problem. Please enter your email address…"}]}]}

That is a well-formed, single-level ADF document and renders fine on its own.

2. The widget then re-wraps it before rendering. In https://jsd-widget.atlassian.com/assets/iframe.js, the bot-message conversion path takes the already-complete ADF document and passes it through a paragraph builder and then a doc builder:

• The paragraph builder converts strings into text nodes and passes objects through unchanged
• The doc builder wraps whatever it receives in {type:"doc", version:1, content:[…]}

Given an ADF object (not a string), the result is a document node sitting in an inline position. That is an invalid ADF, so the Atlaskit renderer falls back to unsupportedInline, whose label is built as "Unsupported " + node.attrs.originalValue.type → "Unsupported doc."

In other words, the conversion looks like it expects a plain string from the server, but the server is sending an ADF object (contentType: "ADF"). Our read is a contract mismatch between the conversation history API and the widget's bot-message renderer.

3. Two observations that corroborate this:

• Reporter (end-user) messages render fine in the same widget—they take a different branch in the same message router, so they never hit this wrapper.
• The portal works because it is a separate application that does not use this widget bundle.

(Module/symbol names are minified and change per build, so we've described the code functionally. We can supply the exact extracted functions, the full HAR capture, and screenshots on request.

Impact

Customer-facing. The virtual service agent has been effectively unusable in our embedded widget, and end users see "Unsupported doc" where an answer should be. No configuration change resolves it—see below.

What we have already ruled out

• Our integration: embed script unchanged for 10 months and identical to Atlassian's documented snippet; reproduces in Atlassian's own preview with none of our code present.
• Knowledge base content/linking: the linked Confluence space (TKB1) has 50+ published articles and is correctly linked to TPS. The fallback message—which comes from no article at all—renders as "Unsupported doc" too.
• KB permissions / article formatting: tested across visibility settings and article styles; no effect. The failure is structural, before any content-specific handling.
• CSP / iframe / hosting: the widget loads and functions; only the message body fails to render.

What we're asking

1. Is this a known defect? If so, please share the public issue key so we can watch it and any ETA.
2. Is there a supported workaround—a widget configuration, a version pin, or a flag—that restores correct rendering in the meantime?
3. Separately: Can you confirm whether custom Rovo (Studio) agents are supported in the embeddable widget or are planned to be? Our widget currently surfaces the generic virtual service agent rather than our custom agent, and we've found no supported way to attach a Studio agent to this channel.