Claude Cowork Ships Record a Skill: Screen-Record a Task Once, Get a Reusable AI Skill

The barrier to teaching an AI a task just dropped from writing a prompt to demonstrating the action once.

Tyler Ross · July 27, 2026

On July 21, 2026, Anthropic added Record a Skill to Claude Cowork, lowering the barrier for 'teaching Claude to do something' from writing a prompt to simply demonstrating the action. Users start a recording, perform the task the way they normally would, and narrate each step out loud. Claude then breaks the recording down into a reusable skill, saves it to the skill library, and the next time the same task comes up, it can be called directly instead of walking Claude through it again.How It...

scene-library

That Number in the Weekly Report Looks Off: Fix It or Send It — There's a Step Missing Between the Two

Seeing a strange number, fixing it outright and sending it as-is are both...

scene-library

Pasting 30 Receipts to Claude at Once: If You Accidentally Paste Them Twice, Does the Amount Get Double-Counted?

Whether pasting twice double-counts isn't about how smart Claude is — it's...

plugins

Before Your First Scheduled Task Goes Live, Spend Five Minutes Seeing What It Would Do — Not What It Did

The biggest difference between a scheduled task and doing it manually: a...

Useful Resources

Claude API Status → Model Pricing → Prompt Playground → Token Counter → MCP Servers → LLM Benchmarks → Model Comparison →

Glossary

View All →

Golden Set

A fixed set of real tasks whose correct outputs have already been approved by a human, re-run after every prompt edit or model change so the difference between the old and new output shows whether the change actually helped instead of just feeling better.

A golden set is a small, fixed collection of real tasks — usually twenty to fifty — where someone has already reviewed the output and confirmed it...

intermediate

Anomaly Detection

Asking Claude to comb through a batch of data and flag whatever clearly deviates from the normal range, attaching a possible direction for the cause — this is the act of finding what looks off, and nothing more. It doesn't include judging whether that oddity is actually an error; that judgment gets handed to someone actually capable of making it.

<a href="/en/glossary/core-concepts/anomaly-detection/">Anomaly Detection</a> refers to asking Claude to find items within a batch of data that...

beginner

Chain of Thought

Ask Claude to write out the intermediate reasoning steps within a single reply, instead of jumping straight to the conclusion — you don't just see the answer, you see how it was derived, which means you can also see exactly which step went wrong.

<a href="/en/glossary/core-concepts/chain-of-thought/">Chain of Thought</a> means asking Claude to write out its reasoning step by step within a...

beginner

Weekly Picks

plugins

Record a Skill or Schedule a Task? First Recognize These Are Two Different Questions

A skill answers how, a schedule answers when. Pick wrong and it's not unusable — it just never...

Derek Finch

scene-library

'Professional But Not Too Stiff': Stop Writing It as a Rule, Paste an Old Email Instead

'Professional but not too stiff' doesn't translate into a rule, but a fitting old email can show...

Jake Sorensen

scene-library

Three Months Into Using a Skill: The Real Problem Is How You Roll Back When an Edit Breaks It

The cost of no version control isn't the breakage itself — it's the blind gap afterward, while a...

Cora Mitchell

scheduled-tasks

A Scheduled Skill Just Failed — Should It Retry Itself, or Wake Someone Up?

Not every failure deserves to wake someone at 3 a.m. Fail to tell transient from structural, and...

Derek Finch