The Intern

Context is the most expensive thing we reconstruct every morning: what moved in Slack overnight, which PRs are waiting on a review, which Jira tickets slipped, which emails actually need an answer. The Intern is a small Claude Code workspace that does this reconstruction for you — it reads the five platforms we work in (Slack, Gmail, Drive, Jira/Confluence, GitHub), remembers what previous sessions found, and hands you back what happened, what needs you, and what's still open on your end.

It is personal by design: your connectors, your relevance lens, your voice. Nothing is shared, and nothing is ever written back to any platform.

The three commands

• /setup: a ten-minute interview, once. Who you are at Ulys, what you're working on, what you want prioritized — plus two or three of your real emails so drafts sound like you and not like a language model.
• /sync: sweeps everything you can see — every Slack channel you're in, every Ulysholdings repo, every visible Jira project and Confluence space, your inbox, your Drive — and files the result under state/pulse/.
• /brief: the payload. What happened, what needs your reaction, and your open items carried forward from previous briefs until they're resolved. Run it any morning.

Setup

1. Install the rig:

   rig hub use hub:hmr.renaudin/intern
   

2. Connect your platforms. In Claude Code, run /mcp (or use claude.ai → Settings → Connectors) and connect Slack, Gmail, Google Drive, and Atlassian with your Ulys accounts. For GitHub:

   gh auth login
   

   Connect what you have; the rig skips missing platforms gracefully and you can add them later.

3. Start working:

   cd intern && claude
   

   On first launch it runs /setup automatically. Answer the questions, then run /sync.

What it will never do

The Intern is read-only on every platform: it never sends an email, posts a Slack message, comments on a ticket or a PR, or changes a label. When a reply would help, it drafts one in your voice for you to copy-paste. Everything it learns about you — your profile, your writing samples, every pulse and brief — lives in local files inside the rig folder and is excluded from packaging, so nothing personal travels if the rig is ever re-published.

What's next

Linear is the next connector we'll wire in, and the sweep design makes that a config change rather than a rewrite. If a sweep misses a channel or a source you care about, or surfaces noise, say so in the session — the rig adjusts its own config files.