The AI workspace for your agent skills

The AI workspace for your agent skills

Your team has built useful AI workflows. The prompt that reliably generates a code review checklist from a PR description. The agent configuration that summarises customer calls into structured notes. The system prompt that makes the AI write in your brand voice. The workflow that takes a meeting transcript and produces action items with assignees. They work. They save time. And they're scattered across individual ChatGPT conversation histories, Claude project configs, Cursor rules files, custom GPTs that one person built, and Slack messages where someone shared "the prompt that actually works."

When someone new joins the team, they build their own versions from scratch. When the person who built the good prompt leaves, the prompt leaves with them. When the team standardises on a workflow, there's no single place to put it where everyone can find it and every tool can use it. The skills exist. The sharing infrastructure doesn't.

Fabric gives your team a shared, searchable library of agent skills: prompts, workflows, system instructions, and agent configurations that anyone can find and any LLM or coding agent can pull from.

One library for every skill your team has built

Store agent skills in Fabric as searchable, versioned documents. A skill can be a system prompt, a workflow definition, a set of instructions for a specific task, a template for a recurring agent job, or a full agent configuration with tools and context. Write them in notes and docs with full Markdown support, or upload existing prompt files, configuration documents, and workflow definitions.

AI search finds skills by what they do, not by what they're called. "The prompt for generating release notes from a changelog" finds it. "Every skill related to customer communication" finds the set. "The agent that summarises sales calls" finds the configuration. Describe the task and find the skill.

The library is shared across the team via real-time collaboration in shared spaces. Everyone contributes. Everyone can search. The skills are a team resource, not individual assets locked in personal tool histories.

Pull skills into any LLM or coding agent via MCP

Fabric's MCP server exposes your skills library to any tool that supports the Model Context Protocol. Your coding agent in VS Code, your AI assistant in the terminal, your custom workflow in Claude or ChatGPT, any MCP-compatible client can pull skills from your Fabric library at runtime.

This means the workflow your team perfected lives in Fabric and is usable from whatever tool each person prefers. The developer who uses Cursor pulls the same code review skill that the developer who uses Claude Code pulls. The marketer who uses ChatGPT accesses the same brand voice prompt that the content lead uses in Claude. The skills are centralised. The tools are individual choice.

The API provides programmatic access for custom integrations. Build workflows that fetch the latest version of a skill, pass it to a model, and return the result. The skill library becomes infrastructure, not a collection of bookmarked prompts.

Version-controlled and always current

Prompts evolve. The first version of the code review checklist missed security considerations. Version two added them. Version three refined the output format. In a personal ChatGPT history, version one and version three are indistinguishable. In Fabric, the current version is the one in the library, and the revision history is searchable.

When someone improves a skill, the improvement is available to everyone immediately. No Slack announcement asking people to update their local copy. No version confusion. The library is the single source of truth for how the team's AI workflows are configured.

Annotations let team members add notes to skills: "works best with Claude for long-form output," "add the project context to the system prompt for better results," "deprecated in favour of the v3 version," "pairs well with the meeting summariser skill." The annotations are searchable, so "every skill optimised for Claude" or "deprecated skills" produces the relevant set.

Organised by function, domain, and team

Smart organization tags skills by what they do: code review, content generation, data analysis, customer communication, meeting processing, research synthesis. Skills for engineering cluster separately from skills for marketing. Internal-facing skills sort differently from client-facing ones. The organisation reflects the function, not when someone created the skill.

Create spaces by team, domain, or workflow stage. Engineering skills in one space. Sales skills in another. Shared skills accessible to everyone. The structure matches how your team actually works.

Agents that use your skills library

Agents in Fabric can draw on skills from the library to run recurring workflows. An agent that runs the "weekly project summary" skill every Friday. An agent that applies the "meeting action items" skill to every meeting transcript. An agent that uses the "code review checklist" skill on every new PR. The skills are reusable building blocks, and agents are the automation layer that runs them on schedule.

The combination of a shared skills library and scheduled agents means your team's AI workflows are documented, discoverable, and running without anyone manually invoking them.

Self-writing documentation for your skills

Self-writing docs can produce documentation about your skills library from your team's Slack discussions and meetings. When someone discusses improving a prompt in Slack, the context is captured. When the team agrees on a new workflow in a meeting, the decision is logged. The skills and the reasoning behind them stay connected.

The AI helps you build better skills

The AI assistant works from your full skills library. Ask it to suggest improvements to a prompt based on what's worked in similar skills. Ask it to compare two versions of a skill and identify what changed. Ask it to draft a new skill based on a description of what you need, grounded in the patterns your team has already established. The library of existing skills becomes the training set for building new ones.

Who uses Fabric for agent skills

Developers building and sharing coding agent configurations. Engineering teams standardising AI-assisted workflows across the team. Product teams maintaining skills for user research synthesis and spec generation. Sales teams sharing call summarisation and follow-up workflows. Marketing teams maintaining brand voice and content generation prompts. Startups building AI infrastructure as the team grows. Indie hackers maintaining personal skills libraries across projects. Freelancers building reusable AI workflows for client work.

For the broader team knowledge approach, see team wiki. For keeping documentation current, see docs that write themselves. For preventing knowledge loss, see knowledge retention.

Get started

Give your team's AI skills a home that any tool can pull from. Try Fabric free. See pricing for teams.

FAQs

Can any LLM or coding agent pull skills from Fabric?

Yes. Fabric's MCP server exposes your skills library to any MCP-compatible client: coding agents, AI assistants, custom workflows, and CLI tools.

Can I search skills by what they do?

Yes. AI search finds skills by function and purpose, not just by name.

Can the whole team share and contribute skills?

Yes. Shared spaces with real-time collaboration let everyone contribute to and search from the same skills library.

Are skills version-controlled?

Yes. The current version is always in the library. Revision history is searchable. Everyone uses the latest version.

Can agents run skills on a schedule?

Yes. Agents can apply skills to recurring workflows: weekly summaries, meeting processing, code review checklists.

Can I annotate skills with usage notes?

Yes. Annotations let you add searchable notes about which models work best, gotchas, and pairing recommendations.

Can I access skills programmatically?

Yes. The API provides programmatic access for custom integrations and automated workflows.

Are skills organised by function?

Yes. Smart organization tags by function, domain, and team.

Can the AI help me build better skills?

Yes. The AI assistant suggests improvements based on patterns in your existing library and helps draft new skills.

What happens when someone who built a skill leaves?

The skill stays in the library. The annotations, the version history, and the context from team discussions are all searchable. No knowledge leaves with the person.

Is our data secure?

Yes. Fabric uses AES-256 encryption and is CASA Tier 2 compliant. Your data is never used to train AI models. Skills in team spaces follow your access permissions.