Let Claude post your updates
You know that moment when you finish something and think "I should post an update about this"... and then don't? Yeah. That's the moment this fixes.
MyStatus has a skill for Claude — Anthropic's AI assistant — that turns it into your personal status reporter. You wrap up a chunk of work, say "post a status update," and Claude drafts the update, shows it to you, and posts it to your task the second you approve. You bring the work; the robot does the paperwork.
A few things we got stubborn about, on purpose
- You approve every word. Claude never posts anything you haven't read. Your task is a public record — nothing goes on it without your say-so.
- One key, one task. The API key you hand Claude is scoped to a single task. It can post there and touch nothing else. You never hand a robot the keys to your whole account.
- Receipts on everything. Updates posted this way are stamped "via Claude skill" right on the public page. Your readers always know who did the typing. (The robots get held accountable too. It's only fair.)
- It learns your voice. Edit its drafts and it takes notes — scope, tone, running bits. Give different tasks different personalities if you want. We do.
Set it up (about five minutes)
- Mint a key. Head over to my-status.app/app/keys, mint a key with integration "Claude skill," and scope it to the task you want Claude posting to. Copy it — it's only shown once. (If you don't, no worries, just create another one. They're free!)
- Install the skill. Download the skill, then in Claude go to Settings → Capabilities → Skills and upload the zip. (Claude Code folks: unzip it into
.claude/skills/instead, but you knew that already.) - Connect the posting line. In Claude, add a custom connector with this URL, pasting in your key:
https://api.my-status.app/mcp?key=YOUR_KEYGot multiple tasks? Mint a key for each and comma-separate them right in that URL — one connector, and the server figures out which key goes with which task. - Tell Claude about your task. First time you say "post a status update," Claude will ask for your task's share link and how you want the updates to sound. Answer once; it remembers.
That's it. From then on: do the work, say the words, approve the draft, watch it land on your task with its "via Claude skill" stamp.
Second movie: the git workflow skill
That first skill handles the telling. This one handles the doing — it's the exact discipline Claude follows while building MyStatus itself. We're not handing you anything we don't run in production, on this very app.
Install it and every coding task Claude touches gets the same rules:
- A branch per task, named after the task, cut off a long-lived
claudeintegration branch. Claude never commits straight tomain. And it won't write a line of code until you've named the task — that prompt is blocking on purpose. - A commit per iteration — including the wrong turns. If a bug hunt took six bad theories, the branch shows six commits. No amending, no squashing. The git log tells the true story of the hunt, not a tidied-up victory lap.
- Nothing merges until you've verified it and reported it. Claude hands you a test list that opens with exactly what proof to capture, you test on the real device, and the work isn't "done" until the status update is posted. (See where this is going? The two skills are a matched set — this one keeps the work honest, the first one tells your readers about it.)
Download the skill, then upload the zip in Settings → Capabilities → Skills (Claude Code folks: unzip it into .claude/skills/). No key, no connector, nothing to mint — this one is pure git discipline, and it works on any project, MyStatus-tracked or not.