Skip to content

Skills Catalog

A skill is a reusable, packaged capability you save once and reuse — or share — instead of rebuilding it from scratch. The clearest example today is a Cowork recipe — a multi-step task you capture so the team can rerun it. In Copilot Studio, skill is a precise, narrower term: a specific kind of tool (an embedded pro-code agent), not a catch-all for everything an agent can do. Same instinct everywhere — capture it once, reuse it — but the Studio meaning is specific, so this catalog keeps the two straight.

This page is the catalog — the reusable skills worth building first, grouped by where they live. Each links to the walkthrough that builds or describes it. Think of skills as the durable counterpart to the prompts and walkthroughs in each stage: a prompt is something you run; a skill is something you keep.

Not to be confused with…

"Skill" is an overloaded word. On this site it means a reusable capability (the definition above). It is not:

  • the Skills agent — a first-party agent that helps you find a career stretch assignment;
  • your personal skill level (new / some / experienced) — the slider on the role paths that tailors a journey to your experience; or
  • the Skills feature inside Copilot Studio — a narrower, product-specific thing: registering an existing pro-code agent (Bot Framework or Microsoft 365 Agents SDK) so your Studio agent can call it as a tool. The Studio items in this catalog are tools/actions, not that feature.

When this page says "skill," it always means a reusable capability you save and reuse.

Why skills matter

The difference between a team that uses Copilot and a team that compounds with it is reuse. A Cowork recipe that saved half a day, a Studio tool your agents reuse instead of rebuilding — each is worth far more when it's captured, named, and rerunnable by someone who didn't build it. Skills are how one person's best workflow becomes everyone's default.

Borrow, don't build

The exact feature names and limits move fast — Microsoft owns the authoritative reference. When a detail matters for customer-facing work, go to the source:

Cowork skills — reusable recipes

In Stage 3 · Cowork, a skill is a reusable recipe: a multi-step task captured once — with its name, inputs, and expected output — so anyone on the team can rerun it cold instead of reinventing the prompt. The ones worth capturing first:

  • Recurring weekly digest

    Describe the Monday-morning digest once and let it run every week without re-prompting.

  • Customer-health digest

    Rank your whole book of business worst-first, on a schedule — so you know where to spend the week.

  • Multi-document brief

    A repeatable "stack of documents → one coherent brief" recipe.

  • QBR prep cycle

    Account data → deck → talking points → follow-ups, run as one rerunnable flow.

  • Incident postmortem

    Reconstruct timeline, root cause, and actions from scattered sources into one blameless draft.

  • Build a recipe library

    Capture any great Cowork run as a reusable recipe so it doesn't die with the session.

Autopilots (Scout) skills — capabilities for an always-on agent

In Stage 5 · Autopilots, a skill is the same reusable capability — but handed to an always-on, autonomous agent. Autopilots are Microsoft's category of such agents, and Microsoft Scout is the first one — so the capability is carried out in the background, without you starting it each time.

  • A Cowork recipe is a skill you describe and run on demand.
  • An Autopilot skill is that same instinct equipped onto an agent that's always watching: the capability Scout draws on to do the job when its trigger fires, then report back.

Emerging and gated

Autopilots are a newer, gated capability — Scout is in Frontier private preview — and the way you equip and manage their skills is still rolling out by tenant, license, and region. This lens stays scoped to what exists today and will grow as the capability does. Treat the Microsoft Scout Learn page as the source of truth before any customer-facing claims.

The skills worth equipping onto Scout first — each captured once, then run always-on in the background:

  • Always-on inbox triage

    Triage your inbox and signals into one short daily brief — what needs a reply, a decision, or just knowing — rerun every workday without you starting it.

  • Continuous customer-health watch

    The worst-first account-ranking recipe from Cowork, equipped onto Scout so the at-risk list re-scores itself and never goes stale.

  • Meeting recap & follow-through

    Every meeting ends with a clean recap, owned actions, and drafted follow-ups waiting for your sign-off — on the trigger, every time.

  • Watch deliverables & flag risks

    A standing watch that blocks time for what's due and raises stalled-decision risks before they become blockers.

Cowork recipe → Autopilot skill

Notice the lineage: the customer-health digest is a Cowork recipe you run on demand; the customer-health watch is that same skill equipped onto an always-on agent. Capture it once in Cowork, graduate it to Scout when it's worth running continuously.

Studio skills — a specific kind of tool

In Stage 6 · Copilot Studio, skill is a precise product term — not a synonym for tool. Today a Studio skill is one specific type of tool: an existing pro-code agent, built with the Bot Framework or the Microsoft 365 Agents SDK, that you register so your Studio agent can call it as a tool from a topic (how it works).

As of today, that's the focused, pro-code definition of a Studio skill. The catalog reflects what's available now, so you'll find guidance scoped to this specific capability rather than a broader walkthrough. As skills in Studio evolve, this section will expand alongside them.

Skills vs. tools in Studio

Connector actions, Power Automate flows, MCP tools, prompts, and topics are all tools — the broad set of things an agent can call. A skill is just one of those tool types (the embedded pro-code agent above). So the Studio building blocks in Stage 6 — connector actions, MCP tools, flows, Fabric connections, triggers — are tools, not skills. Build them there; they're the foundation a skill plugs into.

Harden a skill into a governed tool

Here's the practical bridge. A reusable skill you've been describing in Cowork or equipping onto Scout can be hardened on the Studio surface — rebuilt as a governed tool an agent calls against a real system, under policy and monitoring. Same capability; durable, auditable, and shareable. Each card maps a skill you already run to the Studio tool that hardens it:

One skill, three surfaces

The clearest way to see the whole arc is to follow a single skill across all three surfaces. The customer-health lineage walkthrough does exactly that — the same "rank my accounts worst-first by risk" capability as a Cowork recipe you run on demand, a Scout autopilot that re-scores itself, and a hardened, governed Studio tool that acts on the system of record. Pick the surface by the job, not the product.

Where this leads

The arc is the same instinct growing up. A Cowork recipe is a skill you describe in plain language. When a recipe is stable, repeated, and worth making bulletproof, it's ready to graduate into a real capability in Stage 6 · Copilot Studio — built as a tool (a connector action, flow, or MCP tool) an agent calls against a real system, governed and shared.

Next: Stage 3 · Cowork — where a saved recipe becomes a team skill · Stage 5 · Autopilots for equipping an always-on agent like Scout · or Stage 6 · Studio for the tools an agent calls.

Was this page helpful?

Found an issue or want to suggest an improvement? Open a GitHub issue  ·  Star on GitHub