Claude Record a Skill: Teach Claude by Screen-Recording Your Workflow
Anthropic's Record a skill feature lets you demonstrate a task on screen and have Claude Cowork turn it into a reusable skill—no prompt writing required.

What is Claude's "Record a skill" feature?
Record a skill is a new way to build a Claude Skill by showing Claude a task instead of writing instructions for it. You record your screen while completing a workflow, narrate your reasoning out loud, and Claude Cowork turns that demonstration into a reusable skill it can run again later.
Anthropic announced the feature on July 21, and it lives in the + menu of the Claude desktop app for Pro, Max, and Team subscribers. The point is simple: demonstrate the work once, and skip the hardest part of creating a skill—translating an expert's actions and judgment into clean documentation.
How Record a skill works
The flow is short and requires no coding or markdown authoring:
- Open the + menu in the Claude desktop app and choose Record a skill.
- Start recording and perform the task on your screen as you normally would.
- Narrate your decisions as you go—why you click what you click, what to do on exceptions, which values matter.
- Stop the recording. Claude synthesizes the screen activity, clicks, keystrokes, and your voice commentary into a structured skill.
- The new skill lands in your skill library, where you can trigger it again without repeating the walkthrough.
Under the hood, Claude captures screen activity, mouse clicks, keystrokes, and voice, then processes the demonstration into a reusable skill definition. The narration matters as much as the clicks: it's how Claude learns the reasoning behind each step, not just the sequence.
Why this is bigger than a screen recorder
A Claude Skill is a reusable package of task-specific instructions that Claude loads when it's relevant. Until now, building a custom skill meant authoring a SKILL.md file yourself—a structured markdown document with metadata, step-by-step instructions, and optional executable scripts.
That approach gives you control and transparency, but it takes writing, and the structure can feel intimidating to non-technical users. Record a skill inverts it: the skill is authored for you from a demonstration. That lowers the biggest barrier to skill creation and pushes Claude Cowork closer to an "AI employee" that learns your workflows by watching, rather than only from text prompts.
Skill vs. plugin: a quick distinction
It's easy to mix these up:
- A skill is a single capability—one folder with a
SKILL.mdfile that Claude loads when relevant. Think of it as one repeatable procedure. - A plugin bundles multiple skills together with connectors (integrations like Gmail or Slack) and sub-agents into one installable package. Think of it as a full role, with all the tools it needs.
Record a skill produces the first kind: a single, self-contained procedure.
Who gets Record a skill?
Anthropic has confirmed the feature is available on paid tiers only. Here's the plan breakdown:
| Plan | Record a skill access |
|---|---|
| Free | Not available |
| Pro | Available |
| Max | Available |
| Team | Available |
A couple of details worth knowing: Team plans require a minimum of five seats and come in Standard and Premium seat tiers with different usage allowances, while Max plans run on rolling five-hour usage windows without a seat minimum.
Note that Skills in general are broader than recording skills: Anthropic's help docs list Skills as available across Free, Pro, Max, Team, and Enterprise plans (with code execution enabled), and in beta for Claude Code and API users. Record a skill is the newer, recording-based authoring path layered on top—and that's the part currently gated to Pro, Max, and Team.
Good first use cases
Record a skill shines wherever a repeatable, visually obvious procedure is easier to show than to describe:
- Recurring reports — pull the same numbers from the same dashboards each week and format them the same way.
- Data cleanup — rename, reformat, or reconcile files following house rules you'd struggle to write down.
- Onboarding a routine — capture how an experienced teammate handles a workflow, including the exceptions they know by heart.
- App-specific tasks — steps inside complex desktop software where precise scripting would be painful.
The narration is what makes these durable. Saying "if the invoice total is blank, skip the row" out loud teaches Claude an exception that a bare click-track never would.
Limits and open questions
This is a brand-new feature, so temper expectations:
- It's paid-only. Free users can use existing Skills but can't record new ones.
- Recording quality drives skill quality. A rushed, silent walkthrough produces a weaker skill than a deliberate, narrated one.
- Privacy and retention specifics aren't fully documented yet. The launch announcement is the primary source for the recording feature itself; Anthropic's broader Skills and Cowork docs cover the surrounding product but don't yet answer every recording-specific question about what's captured and how long it's kept. If you record work that touches sensitive data, treat that as an open question until Anthropic publishes more detail.
- If it doesn't appear, update the Claude desktop app first and confirm your account is on an eligible plan.
For the launch details, see Anthropic's announcement on X and the Use skills in Claude help article.
Record once, reuse everywhere with your own AI workforce
Record a skill captures the same idea Eigent is built around: teach an agent a workflow once, then let it run reliably on its own. Eigent is an open-source, multi-agent "Cowork" desktop app that turns repeatable tasks into an AI workforce you run locally—so a recorded procedure becomes a reusable, auditable workflow instead of a one-off. If you're systematizing team routines, our review GitHub PRs workflow and the Claude Tag AI teammate post are good next reads. Download Eigent to build your first reusable skill today.
Recent Posts

Eigent Release Notes v1.0.2: Connector Gateway, In-Session Preview & BYOK Controls
Eigent v1.0.2 adds Connector Gateway, in-session previews, BYOK model controls, clipboard image paste, clearer work logs, and reliability upgrades for agents.

Graph Engineering for AI Agents: Beyond Single Feedback Loops
Graph engineering wires many feedback loops into one network so AI agents don't game their own metrics. See why single loops fail and how to fix the topology.

Kimi K3: Moonshot AI's 2.8T Open-Weight Frontier Model for Agentic Coding
Kimi K3 is Moonshot AI's 2.8-trillion-parameter open-weight model — the largest ever released. See its specs, benchmarks, pricing, and what it means for AI agents.