get-confluence-page

name	get-confluence-page
description	Get a Confluence page or folder (title, body, labels, child pages) by page/folder ID or URL, or search by query, and save everything to .tmp for analysis. Use when the user asks for Confluence documentation, wiki pages, folders, or design docs for Digital Waste Tracking.

Get Confluence page

Read-only fetch of Confluence pages, folders, or search results to .tmp/confluence-*.

Parameters

args is one of:

Input type	Example
Page URL	https://eaflood.atlassian.net/wiki/spaces/WTPG/pages/6511037444/...
Folder URL	https://eaflood.atlassian.net/wiki/spaces/WTPG/folder/6483182044
Page/folder ID	6511037444
Search query	release runbook (optional space: WTPG)

Requires ATLASSIAN_USER and ATLASSIAN_TOKEN. See atlassian-credentials.md. If missing, stop and tell the user to set them in ~/.zshenv.

Route the input

Pick one path:

Input contains /folder/     → Folder workflow (step 1)
Input is plain text/query   → Search workflow (step 2), then page or folder workflow
Otherwise                   → Page workflow (step 3)

Steps

1. Folder workflow

mkdir -p .tmp/confluence-folders/<folder-id>

bash .cursor/skills/get-confluence-page/scripts/folder-contents.sh \
  <folder_id_or_url> \
  .tmp/confluence-folders/<folder-id>

Output: .tmp/confluence-folders/<folder-id>/

File / folder	Contents
folder.txt / folder.json	Folder metadata and hierarchy
pages/index.txt	Page list (includes truncation warning if capped)
pages/<id>/page.txt	Full HTML body per page
pages/<id>/page.json	Raw API response per page

Default max 50 pages. Raise with CONFLUENCE_MAX_FOLDER_PAGES=<n>.

2. Search workflow

Only when the user gives a search term, not a URL or ID:

mkdir -p .tmp/confluence-searches

bash .cursor/skills/get-confluence-page/scripts/search.sh "<query>" [space_key] \
  > ".tmp/confluence-searches/<query-slug>.txt"

Use a filesystem-safe slug from the query (e.g. release-runbook). Show results; ask the user which page or folder to fetch; then run step 1 or 3.

Each search gets its own file — searches do not overwrite each other.

3. Page workflow

For a single page URL or ID:

mkdir -p .tmp/confluence-pages/<page-id>

bash .cursor/skills/get-confluence-page/scripts/page.sh <page_id_or_url> full > .tmp/confluence-pages/<page-id>/page.txt
bash .cursor/skills/get-confluence-page/scripts/page.sh <page_id_or_url> json > .tmp/confluence-pages/<page-id>/page.json
bash .cursor/skills/get-confluence-page/scripts/child-pages.sh <page-id> .tmp/confluence-pages/<page-id>

Output: .tmp/confluence-pages/<page-id>/

File / folder	Contents
page.txt / page.json	Primary page
children/index.txt	Child pages fetched
children/<id>/	Full content per child (max 10)

Raise child limit with CONFLUENCE_MAX_CHILDREN=20.

4. Read saved files

If any script fails, stop immediately. Report the exact error. Do not scrape the UI or guess content.

After a successful fetch:

1. Read folder.txt / page.txt and pages/index.txt or children/index.txt first
2. Read bodies under pages/<id>/ or children/<id>/ as needed
3. For folders, read all saved page files when the user needs full context; for large folders, summarise the index and ask which pages to focus on unless the user asked for everything

5. Respond to the caller

• Single page: return full page content from saved files
• Folder: return folder metadata, page index, and full content of relevant pages (or all saved pages if requested)
• Always mention the .tmp/ path so the user can inspect files directly

Scripts

Script	Purpose
page.sh	Fetch one page by ID or URL
folder-contents.sh	Fetch folder metadata and all pages inside it
search.sh	Find pages by title/body text
child-pages.sh	Fetch child pages under a page
resolve-content-id.sh	Parse numeric ID from page or folder URL

Examples

/get-confluence-page https://eaflood.atlassian.net/wiki/spaces/WTPG/folder/6483182044
/get-confluence-page 6511037444
/get-confluence-page POPs validation WTPG

Related skills

• create-confluence-page — publish or update pages (after reading existing content)
• create-release-note-in-confluence — uses this to load the release note template