Reuse work with Skills

Skills are reusable instructions and resources that give Work specialized capabilities for specific tasks. Use them when a task has a known method, checklist, template, output format, or domain-specific procedure: package an approach once and Work applies it whenever the task matches, with the same tone, format, and procedure every time.

Work loads Skills through progressive disclosure in three stages:

1. Discovery: at session start, Work loads only each Skill's name and description (~100 tokens each), just enough to know when it might be relevant.
2. Activation: when a task matches a Skill's description, Work reads the full SKILL.md instructions into context.
3. Execution: Work follows the instructions, optionally loading referenced files or running examples as needed.

This means you can have many Skills enabled without overloading context: only the relevant ones expand into full instructions.

Vibe Work ships with a set of built-in Skills covering common productivity tasks. They appear in your Skills list alongside personal and workspace ones, activate automatically when a task matches, and can be invoked explicitly with /{skill-name}. Disable any you don't need from Context > Skills.

To see a built-in Skill in action, trigger deep-research explicitly:

/deep-research What are the main approaches to long-context LLMs in 2026?

Work plans the search, runs multiple web queries, synthesizes the sources, and returns a structured brief with citations, opened in Canvas.

Open the sidebar and go to Context > Skills to see your Skills, organized in two more sections:

• Personal: Skills you've created for your own use. Only you see them.
• Workspace: Skills shared across your workspace. Anyone in the workspace can use them.

Toggle each Skill on or off individually to control which ones load into Work's context.

Once a Skill is enabled, Work uses it automatically when the task matches its description. You can also invoke a Skill explicitly:

• Slash command: type / in the chat window and pick a Skill from the list, or type /{skill-name} directly.
• Natural reference: mention the Skill by name in your prompt (for example, "Use the contract-review Skill on this PDF").

When a Skill is active in a task, Work surfaces it in the chat so you know which one is being applied.

There are two ways to create a Skill: build it manually in the editor, or have Work create one for you from a chat.

1. Open Context > Skills in the sidebar.
2. Click New Skill.
3. Fill in the form:
   • Title: a short, descriptive name (for example, Brand voice copywriting).
   • Description: when Work should use this Skill. This text is what Work reads at discovery to decide whether to activate the Skill, so make it specific.
   • SKILL.md: the full instructions Work follows when the Skill activates. Write them as you would brief a colleague: tone, format, steps, examples.
4. (Optional) Add files or folders alongside the Skill (reference documents, templates, brand guidelines, sample outputs, datasets). Work loads them as needed when the Skill activates.
5. Click Create Skill.

The fastest way to build a Skill is to let Work draft it for you from a real task:

1. Run a task where you refine an approach you'd like to reuse (a writing style, a review checklist, a research method).
2. Ask Work to "turn this into a Skill" or "save this as a Skill".
3. Work generates a SKILL.md draft, lists the contents, and proposes it for review.
4. Edit if needed, then save.

From the Context > Skills page, you can:

• Edit a Skill's title, description, SKILL.md, or attached files.
• Disable a Skill without deleting it.
• Delete a personal Skill (workspace Skills require the appropriate permission).
• Publish a personal Skill to your workspace so teammates can use it.

When you edit a Skill, your changes take effect in new chats from that point on. Active chats keep the previous version until you start a new one.

• Build Skills from real tasks. See Create a Skill from a task above. It's the highest-leverage way to start.
• Write descriptions as triggers. The description tells Work when to activate the Skill. "Use when..." phrasing beats "This Skill helps with..." every time.
• Keep Skills focused. One Skill per job. If a Skill grows into many unrelated procedures, split it.
• Avoid near-duplicates. Two Skills with overlapping descriptions confuse Work and inflate context. Edit the existing Skill rather than saving a slight variant.
• Bundle supporting files. Templates, brand guides, sample outputs, or reference data live next to the SKILL.md and load when the Skill activates.
• Iterate. Run the Skill on real tasks, note what's missing, edit, and improve.