A composable library of attachable specialist skills for agent workflows.
Amplifiers is an open-source library of reusable specialist skills that you can attach directly to a task.
The repository is organized around how the library is actually used in daily work:
- attach one specialist skill to a task
- attach multiple skills when the task spans multiple disciplines
- reuse named stacks when the same combinations happen repeatedly
This repository writes skills in the Agent Skills package style used across the broader ecosystem, including OpenAI Codex skills and Anthropic's Claude skill guidance.
Amplifiers now follows four primary top-level layers:
skills/stores the canonical attachable runtime skills.stacks/documents recurring combinations of skills for common deliveries.knowledge/stores internal markdown documentation about the repository.docs/is reserved for the Docusaurus documentation site published via GH Pages.
Deep supporting material now lives inside each skill package, next to the runtime file:
scripts/for executable helpers when a skill needs themreferences/for documentation loaded as neededassets/for templates and packaged resourcesagents/openai.yamlfor optional OpenAI/Codex-specific metadata
Skills are named after the specialist or effect you want to attach.
Examples:
designerhumanizersales-copywriterreact-architectdopamine-driven-copywritter
Stacks are named after the recurring delivery type:
landing-pagelanding-page-reactfrontend-featurerepository-docs
| Skill | Type | Description |
|---|---|---|
| designer | specialist | Design interfaces, pages, and visual systems with hierarchy, UX judgment, and implementation awareness |
| humanizer | modifier | Remove AI writing patterns and restore natural voice, rhythm, and personality |
| sales-copywriter | specialist | Write conversion-focused sales copy for offers, pages, emails, and campaigns |
| dopamine-driven-copywritter | modifier | Increase rhythm, curiosity, and tension in copy without collapsing into hype |
| prompt-engineer | specialist | Turn rough ideas into optimized prompts for specific AI platforms and media types |
| knowledge-writer | specialist | Map a codebase into modular, factual, navigable documentation |
| react-architect | specialist | Define component boundaries, hooks, services, and architecture patterns for React projects |
| django-architect | specialist | Structure Django and DRF backends with clear layers, selectors, services, and view rules |
| laravel-architect | specialist | Structure Laravel backends with clean controllers, services, requests, resources, and integrations |
| html-architect | specialist | Structure semantic, maintainable, accessible HTML projects and marketing pages |
| using-amplifiers | meta-skill | Explain what is available in the library and how to combine it |
| writing-amplifiers | meta-skill | Explain how to create or update skills and stacks in this repository |
| Stack | Description | Uses |
|---|---|---|
| frontend-feature | Ship product-facing frontend work with design, React architecture, and writing polish | designer, react-architect, humanizer, knowledge-writer |
| landing-page | Build a landing page with copy, design, semantic structure, and prompt support | sales-copywriter, designer, html-architect, humanizer, prompt-engineer |
| landing-page-react | Build a React-based landing page with conversion design, sales copy, and frontend structure | designer, sales-copywriter, humanizer, react-architect |
| repository-docs | Improve repository docs, standards, and public-facing structure | knowledge-writer, humanizer, html-architect, designer |
| publication | Distill source material into reusable published skills and stacks | knowledge-writer, prompt-engineer, humanizer |
Browse the repository by layer:
ls skills/
ls stacks/
ls knowledge/Use a skill directly by attaching its SKILL.md to the task context.
Example:
my-skill/
├── SKILL.md
├── scripts/
├── references/
├── assets/
└── agents/
└── openai.yaml
If you are not sure which skill or combination to use, start with using-amplifiers.
.agents/README.mdpoints Codex-style runtimes to the canonical runtime skills inskills/.- Claude Code installation is handled through
.claude/INSTALL.mdand.claude/install_skills.sh.
The Docusaurus site in docs/ is built and deployed by GitHub Actions:
.github/workflows/test-docs.ymlvalidates the build on pull requests.github/workflows/deploy-docs.ymlpublishesdocs/buildto GitHub Pages on pushes tomaindocs/.nvmrcpins the Node version used locally and in CI
Amplifiers follows the same skill-package anatomy documented by Agent Skills, Anthropic, and OpenAI Codex:
my-skill/
├── SKILL.md
├── scripts/
├── references/
├── assets/
└── agents/
└── openai.yaml
In this structure:
SKILL.mdis the required runtime entrypointscripts/is optional executable codereferences/is optional deep documentationassets/is optional templates and packaged resourcesagents/openai.yamlis optional metadata for OpenAI/Codex tooling
The main authoring flow is:
- Published runtime behavior lives in
skills/. - Deep supporting references stay inside the relevant skill package.
- Recurring combinations live in
stacks/. - Internal repository docs live in
knowledge/. - Public site output lives in
docs/.
Follow WRITING_STANDARD.md before publishing new content.
MIT. Use freely, contribute openly.