If you’re curious about my AI commands, I’ll share some details with you. Keep in mind this command was created just by using prompts.
The /start command
This command begins work on a task based on a Trello card. The AI reads the card, determines the task’s name, and asks whether the card has enough detail to start designing.
Here’s the prompt used:
Create a new command called “start” that includes a URL to a Trello card.
Rules:
– If I don’t give a card URL, ask for one.
– Ticket prefix comes from the repo name: <e.g. “ABC” for repo my-app>.
– Ticket number is the number at the start of the URL slug (e.g. …/c/7GVxGJht/78-some-title → 78). If the URL has no slug, use the card’s short ID; if you can’t find one, ask.
– Task type comes from the card’s labels: New Feature → feature, Update → update, Bug → bug, Chore → chore. Ignore other labels. If there’s none of these, or more than one, ask me.
– Task label = <type>/<PREFIX>-<number> (e.g. update/ABC-78). It doubles as the branch name.
– Don’t present a design during /start.Steps:
1. Read the card with the Trello tools: title, description, list, labels, checklists, comments and attachment names.
2. Work out the task label.
3. Check the repo without changing anything: current branch and uncommitted changes. Warn me if I’m not on the main branch, if there are uncommitted changes, or if a local or remote branch with that label already exists.
4. Report briefly: the task label, card title and link; the card’s list, labels and a short summary; whether the card has enough to start designing (if not, numbered questions); any warnings. Don’t analyse the code yet.
5. Switch to plan mode and ask whether the card is enough or I want to add more.Once I confirm, investigate the code and propose a design without changing anything. We refine it together, and you implement only when I explicitly approve.
After creating the file, tell me how to use it.
The takeaway
Once a command is created, you can update it as needed.