A public repo of 12 free example skills published by the team behind HyperAgent, Airtable's always-on AI agent platform. Each skill is a single JSON file holding the reusable context and instructions an agent loads at runtime, so you do not re-explain your brand, method, or format on every prompt. Worth a look as a concrete reference for how a polished agent-skills library is structured, especially since the "skill" idea maps directly onto how I already build with Claude Code.
What it is for
HyperAgent is Airtable's cloud-native agent platform: each session runs in its own compute environment, browses the web, makes decisions, and produces outputs, with hooks into Airtable, Slack, Gmail, and Google Calendar. A skill is the operational knowledge an agent reuses across runs. Not just docs, but instructions the agent follows; when you hit a new edge case you update the skill once and the next run improves. The same pattern as Claude Code skills.
This repo is the team's own set of worked examples, free to read, copy, and adapt.
Format and use
- Language/format: plain JSON, one file per skill (
skill-*.json), roughly 22 to 70 KB each. - No install. Nothing to run locally. Open a file to study how a skill is composed, or import it into HyperAgent.
- No license set on the repo as of June 2026, so treat it as reference material rather than something to redistribute wholesale.
- Last updated 12th June 2026, 376 stars, 27 forks.
The 12 skills
A spread across design systems, content production, and ops:
| Skill | Domain |
|---|---|
| Airtable Kanban Work Tracker | Ops / project tracking |
| Business Simulation Operator Method | Ops / strategy |
| Landscaping Design and Quote | Vertical workflow |
| Brand Book Generator | Brand / design |
| Müller Brockmann Grid Systems | Design system |
| Vignelli Canon Design System | Design system |
| NYT Data Visualization | Data viz |
| NYC Subway Campaign | Design / campaign |
| Briefing Trailer | Video / content |
| Claymation Explainer | Video / content |
| Claymation Podcast Clips | Video / content |
| Veo Hyperframes | Video / generative |
The design-system skills (Müller-Brockmann, Vignelli) are the interesting ones: they encode a named visual methodology so the agent produces on-brand output without re-prompting.
Why it matters for me
I build my own skills (the write-note skill, sales tooling) and the concept here is identical: capture repeated context once, version it, let the agent load it each run. This repo is a free, well-made corpus to study for structure and prompt patterns, and a reminder that the skill is the unit of leverage, not the one-off prompt. Related: Karpathy Claude Code Skills, Writing AI Loops.
Further Reading
The explainer on what agent skills are and when to write one:
