plan-from-confluence

plan-from-confluence

Fetch a Confluence page and its full child hierarchy, then feed each page into /speckit.plan one at a time.

Arguments

$ARGUMENTS contains: <confluence-url> [additional-input]

• confluence-url (required): URL to a Confluence page, e.g. https://mycompany.atlassian.net/wiki/spaces/PROJ/pages/123456789/Page+Title
• additional-input (optional): Extra context or instructions to pass as a separate invocation of /speckit.plan

Steps

1. Parse Arguments

Extract the Confluence URL (first whitespace-delimited token from $ARGUMENTS) and any additional input (everything after it).

2. Parse the Confluence URL

From the URL, extract:

• spaceKey: the path segment following /spaces/, e.g. PROJ
• pageId: the numeric path segment following /pages/, e.g. 123456789

Example URL structure: https://<domain>/wiki/spaces/<spaceKey>/pages/<pageId>/<optional-title>

3. Fetch the Root Page

IMPORTANT: Only perform read operations against Confluence. Never create, edit, or delete pages, comments, or any other Confluence resources.

Using the Atlassian MCP server, fetch the root page content by pageId. Retrieve the full page body (storage or view format).

4. Collect All Pages in the Hierarchy

Recursively collect all descendant pages under the root page:

1. Fetch the direct children of the root page using its pageId.
2. For each child page, fetch its full content.
3. Repeat recursively for each child's children until no further descendants exist.

Collect all pages (root + all descendants) as an ordered list, traversed depth-first (parent before its children).

5. Feed Each Page into /speckit.plan

For each page in the collected list, compose a context block and invoke /speckit.plan with it — one invocation per page:

Confluence Page: <pageId>
Space: <spaceKey>
Title: <page title>
URL: <page URL or canonical link>

Content:
<full page body>

Invoke /speckit.plan once per page, passing its context block. Do not batch multiple pages into a single invocation — each page must be a separate call.

6. Pass Additional Input (if provided)

If additional-input was supplied in $ARGUMENTS, make one final invocation of /speckit.plan with that content as the context. This is a separate call from the per-page invocations above.