Antigravity Skill Vault

1What it does

The largest Antigravity-first skill vault — 300+ SKILL.md packages organized as a curated collection with an installer, a CATALOG.md index, and its own management plugin under plugins/. The skills are ported from the Claude Code agents ecosystem and unified into the Antigravity skill format, covering three shapes of content: domain skills (python-pro, fastapi-pro, async-python-patterns, typescript-pro, rust-pro, k8s-manifest-generator), specialist agent personas (backend-architect, security-auditor), and command or workflow conversions (full-stack-orchestration, conductor-implement). Everything is a directory with a SKILL.md — readable, reviewable, and selectively installable rather than all-or-nothing. At roughly 2k stars it is the most-starred Antigravity-native skill repository, and the MIT license makes it fork-friendly for team-internal vaults.

2When to use it

Reach for it as your first skills install — the catalog is broad enough that most stacks find a dozen relevant skills, and the porting work (tool names, paths, invocation conventions rewritten for Antigravity) is done rather than left as an exercise. It fits developers who want breadth — one vault covering languages, ops, security and business skills — over depth in a single domain. Skip it, or cherry-pick from CATALOG.md, if you only need two or three skills: installing the whole vault means reviewing 300 packages of executable trust, and the honest answer for most projects is to enable the 10-15 that match your stack and leave the rest dormant. Teams standardizing on curated subsets should read the security model in SECURITY.md first.

3How to get started

Clone the repository and browse CATALOG.md first — it indexes the skills by category (development and languages, operations, security, business) so you can shortlist before installing anything. Skills install as folders under your workspace .agents/skills/ or the global Antigravity skills directory — copy the directories you want or use the repo's own bin//scripts/ helpers, which wrap the same folder operations. Each skill is self-contained: the SKILL.md at its root carries the frontmatter the agent matches on, plus any bundled scripts and reference files it needs. After placing a skill, verify discovery by asking the agent what skills it sees, or run a task that matches the skill's description and watch it activate.

4Configuration

The vault's own management plugin (plugins/antigravity-skills- manager) handles install/list operations if you prefer the plugin path over manual copying. Per-skill configuration lives inside each SKILL.md — read the frontmatter description before enabling a skill, because a broad description activates in more contexts than you expect. Bundled scripts inside a skill run with the agent's permissions; the SECURITY.md policy documents what the vault does and does not execute, and the same audit is worth doing on any skill you enable. For team installs, keep the vault read-only and copy into a shared .agents/skills/ — editing inside the cloned vault makes upstream pulls painful.

5Things to know

• Ported, not authored from scratch — the lineage is the Claude Code agents repository, and the port quality (Antigravity-native instructions, not leftovers) is the differentiator to check on the specific skills you enable.
• 300+ skills is discovery surface, not a recommendation to install all — dormant skills cost nothing, but enabled ones compete for activation; keep the enabled set small and specific.
• Skills that bundle scripts are executable trust — the vault ships a security policy, which puts it ahead of most collections, but treat scripts/ folders like dependencies.
• Active maintenance: ~70 commits and a CHANGELOG, which matters in an ecosystem where skill formats are still settling. - Maintenance cadence is visible in commits — a vault this size rots unevenly, so check the last-commit dates on the specific skills you enable rather than trusting the repo-level activity.
• The skills/ tree is browsable directly — for teams that cannot run installers, a straight copy of the folders you want into .agents/skills/ is a fully supported path.
• Activation matching is description-driven — a skill fires when its description matches your request, so skills with broad descriptions will surface often; renaming or narrowing the description of a chatty skill is a legitimate local edit.

6Related on the map

References

• Repository
• Antigravity skills documentation