A planning document for building a streamlined text-to-kanban pipeline using AI

The goal of this project is to create a workflow where a plain-text or Markdown file (maintained by a human or AI assistant like Claude) serves as the source of truth for tasks, which can then be pushed into a Kanban board tool automatically or semi-automatically — reducing manual data entry and keeping project management frictionless.

When working with AI coding assistants like Claude Code, tasks and to-dos naturally surface in text files (CHANGELOGs, TODO.md, notes). Rather than manually re-entering these into a Kanban tool, the vision is to let an AI skill parse and sync them directly to a board — ideally Nextcloud Deck, which is already in use and self-hosted.

Recommendation: Use Nextcloud Deck as the Kanban target and a Markdown file (in DokuWiki or a Claude Code project) as the source.

A structured Markdown file with inline metadata provides a human-readable and AI-parseable task list. Example:

## TODO
 
- [ ] Fix login redirect bug | priority:high | label:backend
- [ ] Write onboarding docs  | priority:low  | label:docs | due:2026-03-20
- [ ] Review PR from Sam     | label:review
 
## IN PROGRESS
 
- [ ] Refactor auth module | label:backend
 
## DONE
 
- [x] Deploy staging environment

This format is compatible with DokuWiki pages, GitHub, Obsidian, and plain editors. Section headings map directly to Kanban columns.

1. Developer or AI edits a .md or DokuWiki task page
2. Claude (or a script) reads the file and extracts structured task JSON
3. An adapter script calls the Nextcloud Deck REST API
4. Cards appear on the Kanban board automatically

[Text file / DokuWiki page]
         |
         v
[Claude parses tasks → JSON]
         |
         v
[Adapter calls Nextcloud Deck API]
         |
         v
[Cards created on Kanban board]

A future Claude Code skill (e.g. sync-tasks-to-deck) would consist of:

• Parser module — reads Markdown task format, outputs structured JSON
• Target adapter — handles Nextcloud Deck API authentication and card creation
• Sync strategy config — create-only, update-if-exists, or archive completed
• Column mapping — maps Markdown headings (TODO / IN PROGRESS / DONE) to board columns
• Label/priority mapping — maps inline metadata to Deck labels and card properties

Invocation example from Claude Code:

"sync my CHANGELOG.md tasks to Nextcloud Deck"

1. Where should the canonical source file live — DokuWiki, a Claude Code project, or both?
2. Should sync be one-way (text → board) or bidirectional?
3. How to handle task identity/deduplication across syncs?
4. Should completed tasks ('DONE') be archived or deleted on the board?
5. Authentication: how to securely store the Nextcloud API token for the skill?

1. [ ] Decide on source file location
2. [ ] Test Nextcloud Deck API with a simple curl request
3. [ ] Draft the JSON schema for parsed tasks
4. [ ] Prototype a parser in Python or a Claude Code skill
5. [ ] Evaluate DokuWiki todo plugin as a possible UI layer for the source file

• Nextcloud Deck API Docs
• Trello REST API
• DokuWiki Todo Plugin
• Claude Code Documentation

kanban ai-workflow nextcloud dokuwiki claude-code automation