FLINT MARKDOWN EXPORT · USE CASE
Prepare selected Confluence pages as context for an AI assistant
Package a small set of project pages as Markdown an assistant can read, and know exactly what you are handing over.
Finished result
A four-page project brief exported as Markdown, with plain-text panels, a combined all-pages.md of 2,436 bytes and one recorded warning. You will know which parts the combined file carries on its own and which still depend on the adjacent files.
Flint Markdown Export does not send content to an AI provider, and it has no built-in AI integration. It prepares files in your browser, or in Atlassian-hosted functions for a background export. Choosing a destination, and following your organization’s rules for what may be shared with it, is a separate step that you control.
Example: a project brief
The fictional Harbor Lights brief (page 2001) has three child pages: Scope and goals (2002) with a flow diagram, Decision log (2003) with a dated table, and Open questions (2004) with a task list, a Jira issue key and a Jira issue list. The brief also links to a Visitor policy page that is not part of the export.
Export the brief
- Open the brief and choose Apps → Export to Markdown, then This page and descendants. Keep the four pages selected.
- Choose AI context. It uses relative Markdown links, filenames with titles and page IDs, and plain-text panels.
- Decide what metadata travels with each page. Open Advanced settings and clear any Frontmatter fields the destination does not need. Clearing
authorandlast_updated_bykeeps those names out of page frontmatter, but they remain inmanifest.jsonand any @mentions in the text stay as names. - Choose Review selection → Prepare full preview. In Preview file, pick
all-pages.mdand switch between Rendered and Markdown source. Open Open questions to see its conversion warning. - Choose Build ZIP from preview → Download ZIP.
What the export contains
harbor-lights-project-brief-2001/
harbor-lights-project-brief-2001.md
harbor-lights-project-brief-2001/
scope-and-goals-2002.md
scope-and-goals-2002.assets/check-in-flow.svg
decision-log-2003.md
open-questions-2004.md
all-pages.md
manifest.jsonOne warning is recorded for Open questions: Jira issue list not exported (JQL: project = HL AND statusCategory != Done). The issue key HL-12 on the same page becomes an ordinary link to Jira; the live issue list becomes an HTML comment marker. If the assistant needs those issues, add a summary to the page or provide the list separately.
The combined file
all-pages.md joins the selected pages in tree order. Each page starts with a comment, its title and a source line, but the YAML frontmatter appears only in the individual files:
<!-- page 2001: harbor-lights-project-brief-2001.md -->
# Harbor Lights project brief
Source: https://example.atlassian.net/wiki/spaces/PRJ/pages/2001/Harbor+Lights+project+brief · Updated: 2026-10-08
...
Figures in this brief are planning estimates, not approved budget.
Background reading: the [visitor policy](https://example.atlassian.net/wiki/spaces/PRJ/pages/2900/Visitor+policy) and the [Scope and goals](harbor-lights-project-brief-2001/scope-and-goals-2002.md) page.Three details in that excerpt matter for an assistant:
- The note panel became an ordinary sentence. With plain-text panels, nothing marks it as a caveat. If emphasis matters, change Panels to Blockquotes in Advanced settings.
- The Visitor policy link still points to Confluence because that page was not selected. A destination without Confluence access cannot follow it.
- The Scope and goals link and the diagram (
) are paths to other files in the ZIP. Uploading onlyall-pages.mdsends the text, not the image or the separate page files.
Trimmed frontmatter
With title, page_id, source_url, labels and updated kept, the brief’s file starts like this:
---
title: "Harbor Lights project brief"
page_id: "2001"
source_url: "https://example.atlassian.net/wiki/spaces/PRJ/pages/2001/Harbor+Lights+project+brief"
labels: ["project-brief", "harbor-lights"]
updated: "2026-10-08T15:30:00.000Z"
---
Match the files to the destination
Check the destination’s own documentation for the file types, number of files and sizes it accepts, and whether it reads folders, ZIP archives or images. Then choose:
- One text file accepted: use
all-pages.md(2,436 bytes, 311 words in the sample). Expect links to other files and images to be paths only. - Several files accepted: provide the individual page files with their
.assetsfolders, so each page keeps its own metadata and images. - Neither fits: narrow the selection or split the brief into topics rather than trimming the Markdown by hand.
Sample files
- PRJ-harbor-lights-project-brief-2001-2026-11-03.zipThe complete AI context export.
- all-pages.mdThe combined file on its own, for inspection.
- manifest.jsonPage paths, metadata, options and the recorded warning.
Limits that affect this task
- Only pages you can read are exported. The files do not keep Confluence permissions once they leave the app.
- The export is a snapshot. Re-export when the brief changes; the assistant does not see later edits.
- Macros that Flint cannot convert or render leave markers and warnings. Read how to fix export warnings before relying on the content.
Sample files on this page are fictional. They were produced by the released Flint Markdown Export exporter, using a simulated Confluence site instead of a live one. Interface labels come from the released app; this guide contains no app screenshots.