What Is a "Skill," Really?
A plain-English primer for lawyers — whether you've been running AI workflows for a year or you opened your first legal AI tool last week.
The one-sentence version
A skill is a reusable set of instructions that teaches an AI how to do a specific task the way you want it done — so you don't have to re-explain yourself every time.
That's it. Everything below is detail.
The analogy that actually lands for lawyers
Think about onboarding a new associate. You don't want to walk them through "how we review an NDA" every single time. So you hand them a checklist: what to look for, the clauses that always matter, your firm's fallback positions, the format the partner expects.
A skill is that checklist — except the AI follows it exactly, every time, without drift, fatigue, or a Friday-afternoon shortcut.
A one-off prompt is you explaining the task once, out loud, to a temp. A skill is the written playbook you keep on the shelf and reuse for years.
Prompt vs. skill: the difference that matters
| A prompt | A skill | |
|---|---|---|
| Lifespan | One conversation | Reused indefinitely |
| Consistency | Varies with how you phrase it | Same every run |
| Shareable? | You'd copy-paste it | Packaged to hand off |
| Captures know-how? | Lives in your head | Lives in the file |
When you find yourself typing the same elaborate instructions into an AI tool for the third time, you've found a skill waiting to be written.
What's actually inside a skill
At minimum, a skill is a written set of instructions. Better skills also carry supporting material the AI should lean on:
- The task definition — what it's for and when to use it
- The method — the steps, the order, the logic ("if the contract is governed by English law, then…")
- Your standards — firm fallback positions, preferred phrasing, formatting, tone
- Reference material — a clause bank, a playbook, example outputs, a checklist
The good ones encode judgment, not just steps. That's what makes them worth sharing.
The two flavors of skill (this is the important part)
Not all "skills" are the same thing under the hood, and the difference decides whether you can move one between tools.
1. Portable skills
These are self-contained files — most commonly a format called a SKILL.md — that you can install into a compatible AI tool, hand to a colleague, or publish to a marketplace. Download, drop in, run. They travel.
This is the world where "import" and "export" are literal buttons. It grew up mostly around general-purpose AI assistants and developer tooling, and it's where the open skill marketplaces live.
2. Platform-native skills
These live inside one vendor's product — Harvey's Workflow agents, Legora's Workflows and Playbooks, and similar features in other vertical legal AI tools. They're often more powerful inside their home turf, because they're wired directly into that platform's documents, citations, and review tools.
The trade-off: they don't travel. You build them in the platform, you share them within your firm's account, and that's the boundary. There's no standard file you export and import into a different vendor — at least not today.
Why this matters for you: if someone says "I'll send you my due-diligence skill," the first question is which flavor. A portable skill you can install. A platform-native one, you'll typically rebuild from the shared instructions inside your own tool. Same know-how, different delivery.
Why lawyers should care at all
- Consistency — every matter gets the same rigor, not whatever the AI improvised that morning.
- Captured expertise — the senior partner's review method stops living only in the senior partner's head.
- Faster onboarding — a new joiner runs the firm's skills on day one instead of learning them over months.
- Less wasted time — you stop re-typing the same setup and start from a known-good baseline.
- Defensibility — a documented, repeatable method is easier to stand behind than an ad-hoc prompt.
The honest caveat
A skill makes the AI consistent, not correct. It will faithfully apply a flawed method just as reliably as a good one. Skills are a way to scale your judgment — they are not a substitute for it. Every output still needs a lawyer's review before it reaches a client or a court. That's not a limitation of the technology; it's the job.
Where Skillagram fits
Skillagram is where you discover skills, share the ones you've built, and figure out how to bring them into the tool you actually work in. Some you'll install directly. Others you'll rebuild inside your platform from the instructions — and we'll show you how for each one.
New to this? Start by spotting one task you've explained to an AI more than twice. That's your first skill.