Clone this repo into your Claude Code skills directory:
```
git clone <repo-url> ~/.agents/skills/planka-skill
```
Or copy the folder manually to ~/.agents/skills/.
Generate an API key in Planka: User Settings → API Key
Set environment variables (in project .env, ~/.env, or shell):
```
PLANKA_URL=https://planka.example.com
PLANKA_API_KEY=your-api-key
```
A trailing slash in PLANKA_URL is fine. The skill strips it before adding /api. When both env files exist, the project's .env overrides ~/.env; values in either file override existing shell values.

Helper CLI

Run from the target project's root. If the skill lives elsewhere, replace the script path with its installed location. The CLI needs only Python 3's standard library.

python3 .pi/skills/planka/planka.py                         # projects, boards, card counts
python3 .pi/skills/planka/planka.py board TodoApp             # lists and cards
python3 .pi/skills/planka/planka.py card TodoApp "Login Bug"   # card details
python3 .pi/skills/planka/planka.py card TodoApp "Login Bug" --full
python3 .pi/skills/planka/planka.py spec endpoints card-labels
python3 .pi/skills/planka/planka.py spec refresh

After the user approves a board write:

python3 .pi/skills/planka/planka.py create-card TodoApp TODO "Fix login" --description "Steps to reproduce..." --confirm
python3 .pi/skills/planka/planka.py move-card TodoApp "Fix login" "IN PROGRESS" --confirm

The script resolves board, list, and card names to current IDs. Ambiguous names fail instead of picking one. Writes require --confirm and prior user approval in conversation. It refuses unknown flags. Run each subcommand with --help for arguments.

The first API call downloads the published OpenAPI spec to .pi/cache/planka-swagger.json in the current project; later calls use the cache. Run spec refresh to fetch a new copy. The cache is public schema data, not credentials. If the server's API differs from the published spec, refresh and inspect its version rather than guessing paths. A first-time run needs network access to both the spec and Planka; subsequent endpoint checks work offline, but board operations still need Planka.

For unsupported operations, inspect the cached spec first (spec endpoints <term> and the request schema in the JSON file). Use the API base ${PLANKA_URL%/}/api, X-Api-Key authentication, encoded JSON bodies, and validate the response content type and JSON. Never print the key. The helper reads simple KEY=value / quoted values for PLANKA_URL and PLANKA_API_KEY from ~/.env then .env; it does not execute shell expressions in those files.

Quick Commands

/planka overview              # all projects and boards at a glance
/planka board "My Project"    # detailed board view with cards and progress

What You Can Do

Talk naturally — the skill understands intent and figures out the API calls.

Browse & Inspect

/planka show me all my boards
/planka what's in the ToDo list?
/planka show me card "Login Bug" with all its tasks

Create & Organize

/planka create a card "Dark Mode" in Backlog with fitting labels
/planka set up a new board "Backend" with lists ToDo, In Progress, Done, Blocked
/planka add labels Bug (red), Feature (blue), Urgent (red) to the "My Project" board
/planka add tasks to "Dark Mode": update theme provider, test contrast, update docs

Move & Update

/planka move "Dark Mode" to In Progress
/planka mark "Dark Mode" as done
/planka rename "Dark Mode" to "Dark Mode v2"
/planka add a comment to "Dark Mode": waiting for design review

Bulk Operations

/planka move everything from In Progress to Done
/planka read the git log and create cards for all completed features
/planka look at board X and create the same structure for a new board
/planka check all cards for typos and fix them

Project Sync

/planka sync the board with our recent git commits
/planka create cards for all open issues
/planka what changed on the board since last week?

The helper caches the Planka OpenAPI spec. Refresh it explicitly when the API changes.

Requirements

Planka instance with API key support
Python 3