AI skill management for engineering teams

You don't know what your developers are telling AI.

SkillInvoke gives engineering leads a management layer for AI coding skills. Write them in GitHub. Bundle them into packs. Sync them to every machine.

Free for small teams No credit card 2-minute setup
~/.claude/skills/
$ skillinvoke sync
Fetching skill manifest...
Found 12 skills across 3 packs.
 
frontend/react-patterns ··········
frontend/css-standards ···········
backend/rails-conductor ··········
backend/api-standards ············
shared/code-review ··············
... and 7 more
 
Sync complete: 12 installed, 0 unchanged.
 
$

Works with

Claude Code | Cursor | GitHub Copilot | Windsurf

The problem

Every developer writes their own AI skills. Nobody reviews them. Nobody knows what's in them.

Skills scattered across machines

No shared standards. Every developer has their own collection of AI prompts with no consistency.

No review process

Skills change without oversight. Nobody signs off on what the AI is being instructed to do.

Zero visibility for leads

Engineering managers have no idea what instructions their team's AI agents are following.

New joiners start from scratch

Institutional knowledge about how to use AI effectively stays locked in individual machines.

How it works

Three steps to governed AI skills

01

Write SKILL.md files in your repo

No proprietary format. No new language to learn. Each directory with a SKILL.md becomes a skill. Use YAML frontmatter for metadata, Markdown for instructions. Your team already knows how.

  • Works with any AI tool that reads skill files
  • Full git history and version control
  • Easy to write, review, and understand
frontend/react-patterns/SKILL.md
1---
2name: React Patterns
3description: Standard React component patterns
4category: Frontend
5tags: [react, components, hooks]
6user-invocable: true
7---
8
9# React Patterns
10
11When building React components, follow
12these conventions:
13
14## Component Structure
15- Use functional components with hooks
16- Extract custom hooks for shared logic
terminal
$ skillinvoke list
 
PACKS
──────────────────────────────────────────
Frontend Essentials 4 skills [mandatory]
Backend Toolkit 5 skills
Shared Standards 3 skills [mandatory]
 
SKILLS PACK
──────────────────────────────────────────
react-patterns Frontend Essentials
css-standards Frontend Essentials
testing-guide Frontend Essentials
accessibility Frontend Essentials
rails-conductor Backend Toolkit
... 7 more

02

Bundle and govern in SkillInvoke

Group related skills into packs. Mark them as mandatory or optional. Assign by team or role. Edits go through pull requests — skills don't reach developer machines until someone approves them.

  • Mandatory and optional pack policies
  • PR-based review workflow
  • Role-based team assignments

03

Sync to every developer's machine

One command to install. One command to sync. Single Go binary, zero dependencies. Sync globally or per-project — skills land exactly where your AI tools expect them.

  • Install with a single curl command
  • --global for shared skills, --project for repo-specific
  • SHA-verified downloads, secure by default
terminal
# Install the CLI
$ curl -fsSL https://skillinvoke.com/install.sh | sh
Installed to /usr/local/bin/skillinvoke
 
# Authenticate (opens browser)
$ skillinvoke login
Authenticated as acme-corp
 
# Pull skills globally or per-project
$ skillinvoke sync # → ~/.claude/skills/
$ skillinvoke sync --project # → .claude/skills/
 
Sync complete: 12 installed, 0 unchanged.
$

Features

Everything engineering leads need

GitHub-native

Skills live in your repo as SKILL.md files. Full git history. No vendor lock-in. GitHub is the source of truth.

Pack system

Bundle related skills into mandatory or optional packs. Assign by role, team, or project. Define in YAML or manage in the app.

PR workflow

Edits go through pull requests for review. Approve or request changes before skills reach developer machines.

CLI sync

Single Go binary. No runtime dependencies. Install via curl. Sync globally or per-project with one command.

Feedback & analytics

Developers rate skills after use. See what's working, what isn't, and who's contributing the most value.

Audit trail

Every edit, sync, and assignment is logged with full provenance. Export audit logs for compliance reporting.

Pricing

One product. Three plans. No surprises.

Start free. Upgrade when your team grows.

Free

£0

For individuals and small teams getting started.

  • 10 skills
  • 3 team members
  • 1 pack
  • CLI sync

Business

£99 /month

Unlimited everything, plus compliance.

  • Unlimited skills
  • Unlimited team members
  • Everything in Team
  • Audit log exports
  • Priority support

FAQ

Common questions

Why can't I just write scripts or keep a shared folder of prompts?

You can — and most teams start there. The problem comes at scale: who reviews changes to those scripts? How do you enforce that every developer is using the approved version? How does a new joiner get the right set on day one? SkillInvoke adds the governance layer that scripts alone can't provide: PR-based review, mandatory pack policies, role-based distribution, and an audit trail of every change. The skills themselves are still just Markdown files in your repo.

Is this vendor lock-in? What if I stop using SkillInvoke?

No lock-in. Your skills are plain Markdown files stored in your own GitHub repo — we never host or own them. Pack manifests are YAML. If you cancel tomorrow, nothing changes in your repo and the last-synced skills stay on every developer's machine. SkillInvoke is a management layer on top of files you already control.

Isn't this just another abstraction layer I don't need?

SkillInvoke has zero runtime overhead. There's no SDK to import, no wrapper around your AI tool, and no process running in the background. The CLI is a single Go binary that copies files from your repo to the skills directory — that's it. We're closer to a deployment tool than an abstraction layer. Your AI tools read the same SKILL.md files they always would; SkillInvoke just makes sure the right ones get to the right machines.

What AI coding tools does SkillInvoke work with?

Claude Code, Cursor, GitHub Copilot, and Windsurf. Any tool that reads SKILL.md files from the standard skills directory. The format is open and portable — we don't lock you into a specific agent.

Where are my skills actually stored?

In your GitHub repository. SkillInvoke reads SKILL.md files from your repo — it never stores skill content on our servers. GitHub is the source of truth. If you stop using SkillInvoke, your skills stay exactly where they are.

How do developers install the CLI?

Single binary, no dependencies. Run curl -fsSL https://skillinvoke.com/install.sh | sh to install. Then skillinvoke login to authenticate via browser and skillinvoke sync to pull skills. Use --project to sync to the current repo instead of globally. Takes about two minutes.

Can SkillInvoke access or modify my code?

No. We use a GitHub App with read-only access to the specific repository you connect. We read SKILL.md files and pack manifests. We cannot see your source code, push commits, or modify anything in your repo.

Can I use SkillInvoke for free?

Yes. The free plan covers up to 10 skills, 3 team members, and 1 pack — enough for most small teams. No credit card required. Upgrade to Team when you need PR workflows, analytics, and more seats.

What happens if I cancel?

Your skills stay in your GitHub repo — we never touch them. The CLI continues to work with whatever was last synced. Your access to the web dashboard continues until the end of your billing period. No lock-in, no data hostage.

Stop the AI skill free-for-all.

Give your engineering team the skills they need, governed the way you want.

Free forever for small teams. 14-day trial of Team features.