Three shifts in your day.
Status comes from the source
"What actually landed this week?" gets answered from changelists, build results, and closed tickets — not from interrupting six people for estimates of their own progress.
Meetings produce artifacts
Every meeting ends with notes, decisions, and owned action items drafted before people are back at their desks. Follow-up stops leaking.
Plans survive contact with reality
Specs become task breakdowns with dependencies in minutes; when scope shifts, Claude re-cuts the plan and shows you exactly what moved and what it costs.
What you'll run every week.
Production leans less on the Unreal-specific skills and more on Claude's synthesis muscle — but the project-aware moves below are what make our setup different from a generic AI notetaker.
Browse all Production skills in the Library →
You don't run builds — but you live downstream of them. Claude reads build logs and changelists so you always know: is the nightly green, what broke it, who's on it, and which milestone build has which features. The morning standup question answers itself before standup.
The studio's full pre-release gauntlet as an executable checklist. Claude walks it, gathers evidence per item, and hands you the exceptions list — so a ship decision is a review of five open questions, not a hunt through two hundred settled ones.
Hand Claude a design doc and it drafts the work breakdown: engineering, design, art, and QA tasks, dependencies between them, and the questions the spec doesn't answer. Because it can read the codebase, it flags where the spec collides with how systems are actually built — before the collision costs a sprint.
Feed Claude the transcript or your raw notes; get back decisions made, action items with owners, and open threads — formatted for wherever we track work. Ask for the diff against last week's notes and you have your follow-up list too.
Cross-read the bug database, build failures, and ticket churn to answer the questions a producer actually cares about: which system is quietly generating the most rework, which milestone item hasn't moved in two weeks, where the plan and reality are diverging.
The rest of the job gets faster too.
Stakeholder translation
The same status, told three ways: technical detail for leads, milestone framing for partners, one crisp paragraph for investors. Claude re-voices without changing the facts.
Docs, decks & templates
Milestone review decks, onboarding docs, postmortem writeups, vendor briefs — drafted from bullet points in minutes, in the studio's voice.
Schedule math
"If art needs one more week on Chapter 2 and QA is fixed, what actually moves?" Claude walks the dependency chain and shows you the real critical path, not the vibes-based one.
Process changes go in the playbooks.
New definition of done? Changed the ticket template? Update the relevant skill the same hour you announce it. Announcements decay in a week — skills are how process changes actually stick, because every department's Claude starts enforcing them immediately. See exactly how →
Status that comes from the systems, not from asking people.
Jira and Confluence is the big one — Atlassian's server reads and writes issues and pages while respecting the permissions you already have. That's what makes spec-to-plan produce actual tickets instead of a list you then retype, and risk-radar read the real board instead of your memory of it.
Then Slack, so meeting-to-action can reconstruct why a decision was made and whether anyone owned the follow-up. And if we're shipping localized, Crowdin is what stops loc from being an emergency at the end of every milestone. All three, with commands →