AI.Tedt.org AI Library

A unified library of reusable AI assets: Agent Skills, plugins, automations, Agent Instruction Files, Scoped Agent Instruction Files, Agent Definition Files, Prompt Template Files, and Work Specification Files. Every asset uses the same searchable catalog, card system, detail shell, downloads, artwork, ratings, and discussion experience.

Built with Astro + Tailwind CSS, deployed as a static site to GitHub Pages.

Mirror note: This repository is an independently published mirror of Microsoft's CAT Agent Skills repository. Microsoft engagement remains visible as a labeled upstream source, while AI.Tedt.org collects its own ratings and comments in this repository. The upstream Microsoft Clarity project is disabled in this deployment.

Features

• One AI Library with search and filters for asset type, compatibility, contributor, and a controlled 12-topic vocabulary.
• Shared asset pages with a 16:10 cover, metadata, guide, exact source, related assets, downloads, ratings, discussion, and a contextual use panel.
• Prompt customization for text, textarea, number, select, radio, and checkbox variables, with defaults, validation, reset, live preview, download, clipboard copy, and provider launch. Prompt text is never placed in a URL.
• Fresh generated artwork with a validated 1600×1000 WebP and complete generation provenance for every newly published asset.
• Federated community engagement: rate and discuss an asset through GitHub Discussions. Microsoft and AI.Tedt.org activity stays separately attributed, while the gallery uses their combined positive reactions for "Top rated." See docs/ratings.md.
• Exact downloads: individual source files and deterministic ZIP bundles for multi-file assets.
• Stable public interfaces: existing /skills/{slug}/ routes and skills.json remain compatible; the complete catalog is available in assets.json, with prompts at /prompts/{slug}/ and other file assets at /library/{slug}/.

Local Development

npm install
npm run dev      # start the dev server (http://localhost:4321/)
npm run build    # production build into ./dist + metadata/social asset audit
npm run preview  # preview the production build locally

The production site is served from the root of the ai.tedt.org custom domain, so local URLs do not use a repository-name prefix.

Adding an Asset

Add one folder to submissions/ and open a PR. Do not edit generated files in src/content/, src/content/guides/, or public/bundles/ by hand. CI validates the submission and generates its page and downloads. See CONTRIBUTING.md for the full guide.

Generic file submissions declare kind, name, description, tags, author, entrypoint, and explicit payloadPaths in metadata.json or metadata.yaml. The importer supports these generic file kinds:

• agent-instruction
• scoped-agent-instruction
• agent-definition
• work-specification

Optional fields include keywords, compatibility, status, versioning, and provenance. The folder name supplies the slug, and the importer derives the download paths. Filenames and relative payload paths are preserved. Free-form tags remain searchable, while the catalog maps them into the controlled Topics vocabulary below. A root README.md is an optional human-facing guide and is not included in the downloadable payload unless it is explicitly named in payloadPaths.

Topic Taxonomy

The catalog exposes 12 broad Topics. Specific products, formats, industries, methods, and capabilities remain searchable as tags instead of crowding the filter:

• AI and Agents
• Architecture and Engineering
• Business and Strategy
• Communication and Collaboration
• Content and Documentation
• Data and Analytics
• Design and Media
• Industries and Domains
• Governance, Risk, and Compliance
• Knowledge, Learning, and Research
• Planning and Delivery
• Productivity and Automation

For example, a generic Agent Instruction File submission can contain:

submissions/repository-instructions/
├── metadata.json
├── README.md
└── AGENTS.md

{
  "kind": "agent-instruction",
  "name": "Repository Instructions",
  "description": "Persistent operating instructions for a repository.",
  "topics": ["Architecture and Engineering"],
  "tags": ["engineering"],
  "author": "Your Name",
  "entrypoint": "AGENTS.md",
  "payloadPaths": ["AGENTS.md"],
  "compatibility": ["codex", "github-copilot"]
}

Each generic payload is available as an individual source-file download. A multi-file submission also receives a deterministic <slug>.zip containing exactly the declared payloads. A prompt template uses one <slug>.prompt.md entrypoint and may declare customization variables. Existing skill, plugin, and automation submissions continue to work through the compatibility adapter.

Do not add publicationStatus or generated artwork fields to a new prompt or generic file submission. The importer stages it as blocked, and npm run artwork:prepare -- --slug <slug> writes the complete cover record and publishes it atomically after the artwork is ready.

Existing Agent Skill Format

Every submission is a submissions/<slug>/ folder with a metadata.json sidecar plus one skill payload — an unpacked SKILL.md (+ optional dirs):

submissions/<slug>/
├── metadata.json  # OR metadata.yaml — catalog details (sidecar, not bundled)
├── README.md      # OPTIONAL human-facing note (not bundled) — shown on the page
└── an unpacked skill:
    ├── SKILL.md    # name + agent description + instructions
    ├── scripts/    # optional executable code
    ├── references/ # optional docs
    └── assets/     # optional templates / data files

A skill carries two descriptions: the agent description in SKILL.md frontmatter (what the model reads to decide when to invoke), and the catalog description in metadata.json (the friendly one-liner shown in the gallery). An optional README.md adds a third, human-only voice: it never ships to the agent and, when present, becomes the main content on the detail page — the author's overview, setup, and usage — with the exact SKILL.md still offered as the download.

submissions/my-great-skill/SKILL.md:

---
name: my-great-skill
description: Use this skill whenever the user… (the agent-facing trigger).
---

Write the agent instructions here as Markdown — this body becomes the
"Instructions" section on the detail page.

submissions/my-great-skill/metadata.json:

{
  "name": "My Great Skill",
  "description": "One-line catalog summary shown on the card and detail page.",
  "platforms": ["Cowork", "Copilot Studio", "Scout", "Codex"],
  "tags": ["productivity", "automation"],
  "author": "Your Name",
  "authorUrl": "https://example.com",
  "version": "1.0.0"
}

name, description, platforms, and tags are required (platforms must be one or more of Cowork, Copilot Studio, Scout, Codex). PRs run a build check that validates every skill against the schema.

Besides single skills, you can submit a Scout automation (a .json export, Scout-only). It drops into submissions/<slug>/ the same way and is auto-detected by its payload. (Cowork plugins and Scout automation installers were .zip packages and are no longer accepted — .zip payloads have been retired; existing ones stay published.) See submissions/README.md for the details.

Project Structure

src/
  components/      shared catalog cards, detail shell, prompt workbench, and covers
  content/
    skills/        generated skills, plugins, and automations
    prompts/       generated prompt template files
    artifacts/     generated instruction, agent definition, and work specification files
    guides/        optional human-facing submission guides
  layouts/         base page layout and site metadata
  lib/             normalized LibraryAsset model, schemas, engagement, and helpers
  data/            engagement.json + derived ratings.json snapshots
  pages/
    index.astro          unified catalog search, filters, sort, and infinite scroll
    skills/[slug].astro  existing skill, plugin, and automation routes
    prompts/[slug].astro prompt template routes and approved legacy aliases
    library/[slug].astro other instruction, agent, and specification file routes
    tags/[tag].astro     normalized per-topic listings
    social/              generated 1200×630 site and per-asset preview images
    robots.txt.ts        crawler policy + sitemap discovery
    sitemap.xml.ts       canonical, indexable page inventory
    skills.json.ts       backward-compatible skill metadata endpoint
public/
  assets.json      complete published library catalog
  bundles/         exact individual payloads and deterministic ZIP bundles
  skill-art/       validated generated 1600×1000 WebP covers
  site.webmanifest + browser, Apple, and installable-app icons
submissions/       drop-in library submissions (folders imported by CI)
scripts/           import, content validation, metadata audit, icons, engagement
.github/workflows/ CI + Microsoft catalog sync + Pages/engagement refresh

Deployment

Pull requests run .github/workflows/ci.yml, which imports submissions, regenerates assets.json, validates content and artwork, and builds the site. Pushing to main triggers .github/workflows/deploy.yml, which builds and publishes the site to GitHub Pages at https://ai.tedt.org/. Enable Pages in the repo settings with the GitHub Actions source and assign ai.tedt.org as the custom domain.

Before each build, deploy.yml runs npm run engagement:fetch to snapshot Microsoft and AI.Tedt.org reactions, comments, discussion links, and repository stars. Its daily schedule runs at 3:30 a.m. Pacific, after the 3:00 a.m. Microsoft catalog sync, and deploys only when the deterministic snapshots change. See docs/ratings.md for the source and ownership model.

Support

AI.Tedt.org is an independently published community library. Individual assets are provided as-is; provider support policies apply to the tools with which they are used. See SUPPORT.md for how to file issues and get help.

License

MIT

Shared TedT.org Navigation

The global header loads https://tedt.org/assets/global-header/v1/header.js. Menu changes belong in the TedT.org repository, rather than in this site. The small local fallback and pre-paint bootstrap keep navigation and appearance usable while loading. The shared tedt-appearance cookie and tedt:theme-change event synchronize appearance; keep local site navigation and application controls separate. See https://tedt.org/docs/global-header.md for the embed contract.