Open-source AI interfaces built with Claude (Fable 5) including hero sections, GLSL shaders, design systems, animations, 3D components and and landing pages, in React, Tailwind, and Three.js. Plug them into your AI agent and ship faster.
- ✓Open-source license (MIT)
- ✓Actively maintained (<30d)
- ✓Healthy fork ratio
- ✓Clear description
- ✓Topics declared
git clone https://github.com/pulkitxm/claude-directory && cp claude-directory/*.md ~/.claude/agents/4 items in this repository
Use this agent whenever Pulkit gives a new project/UI experiment prompt in the fable repo (a request to build any app, page, scene, component, shader, design system, or UI experiment). The fable repo is a sandbox for experimenting with Fable 5. The agent reads and understands the project prompt, first scans all existing projects across every category to make sure this experiment hasn't already been built (stopping early and reporting the match if it has), reviews the repo's current category folders (today: hero-sections, landing-pages, animations-loaders, 3d-games, portfolios, components-ui, ui-design, shaders — plus any others present on disk), and places the experiment in the best-fitting existing category; it creates a brand-new category folder (with its own README and a matching root-README section) only when none of the current categories fit. Every experiment follows a fixed delivery workflow — worktree → category folder → verbatim Markdown-formatted uppercase prompt.md committed first → full build → CLI-only verification → review → demo.mp4 recording → poster generation → README registration with reconciled counts → commit → PR → merge → branch/worktree cleanup — and the run ends with a consistency sweep that fills any repo-wide gaps (missing demo.mp4/poster.jpg/posters.json entries, drifted counts, typo'd category paths). IMPORTANT - when invoking this agent, pass the user's project prompt VERBATIM (word for word, unmodified) as part of the task, because the agent must preserve it in prompt.md. Do not use this agent for questions or one-off edits to existing experiments.
Use this agent to write (or rewrite) a single project's own `README.md` for SEO and discoverability, after the project has been built. Pass it ONE project folder path (e.g. `hero-sections/aethera-cinematic-hero`); it reads that project's `prompt.md`, `package.json`, source, and demo to produce an accurate, keyword-rich `README.md` with a descriptive H1, a lead paragraph, the real run/verify/demo instructions, and a footer that links back to the category, the root directory, and the live gallery. It never fabricates features or commands — every claim is grounded in the project's own files. The fable-experimenter and template-cloner agents call this agent after their build is verified, to produce the project README before registering the project in the root/category tables. Do not use it for the root README or the category-folder READMEs (those are maintained by the build agents), and do not use it to edit code.
Use this agent whenever Pulkit wants an existing premium-template clone in the fable repo audited against its original source — to confirm the clone really is same-to-same and to FIX it where it drifted. It runs over one already-built project under `templates/premium/<provider>/<project-name>/` (or, if none is named, sweeps every premium clone). For each project it reads the `REFERENCE:` URL from `prompt.md`, re-reconnoiters the live original with `scripts/record-demos/scrape-ref.mjs` (full-page screenshots, computed-style outlines, raw `source.css`, and headless interaction capture), discovers every page the original has, and does the same capture against the local clone — then a vision-judge sub-agent diffs original vs clone across pages present/missing, styles (palette, fonts, type scale, spacing, radii), responsiveness (mobile/tablet/desktop breakpoints), button/hover/focus states, interactive behavior (modals, dialogs, dropdowns, popovers, tooltips, accordions, tabs, menus, carousels, scroll reveals), and light/dark mode. Unlike a report-only check, it APPLIES fixes for every confirmed gap (reusing the cloner's self-correcting vision loop), re-records `demo.mp4` with `record-one.sh`, regenerates the poster, reconciles README counts, and ships it: `make format` → commit → PR → auto-merge to main. Demos serve directly from the repo via the Pages deploy (the old Cloudinary upload workflow was removed), so a merged PR publishes the updated demo automatically — no force-upload step. It also fixes missing pages AND missing/partial sections within existing pages (dropped hero/feature/testimonial/pricing/CTA/footer blocks, abridged card or row counts, placeholder copy), rebuilding them verbatim from the original. IMPORTANT — pass the target project path (or "all") VERBATIM. Use `template-cloner` to build a NEW clone from a URL; use this agent to audit-and-repair clones that already exist.
Use this agent whenever Pulkit gives a UI template/website URL in the fable repo and wants it reproduced same-to-same — every page, the full look & feel, hover states, and scroll/entrance animations — as a self-contained plain HTML/CSS/JS clone. The agent first reconnoiters the source: it crawls the template to discover ALL its pages, then captures a full-page screenshot, computed-style outline (fonts, colors, type scale, spacing, animations), and raw HTML for each page using scripts/record-demos/scrape-ref.mjs. Clones are grouped by template provider under `templates/premium/<provider>/<project-name>/` (e.g. `templates/premium/aceternity/productized-agency/`, `templates/premium/nextjstemplates/sidefolio/`). It checks the reference URL isn't already cloned (under templates/ or the legacy studies/), writes the reference URL + a derived style/layout breakdown into an uppercased prompt.md (committed first), extracts shared design tokens, then fans out one builder sub-agent per discovered page (via a Workflow) to rebuild each page. Each page runs a self-correcting vision loop: rebuild → re-run scrape-ref.mjs against the local clone → a vision-judge sub-agent diffs the clone's screenshot/outline against the reference and returns concrete fixes → apply → repeat until visually faithful or the iteration cap is hit. It finishes with the same delivery tail as fable-experimenter — record-one.sh demo.mp4 → poster → register in root + templates/README.md with counts reconciled from disk → consistency sweep → PR → merge → worktree cleanup. IMPORTANT — pass the user's template URL VERBATIM as part of the task; it is preserved as the REFERENCE in prompt.md. Use this agent for "clone/copy this template/site" requests with a URL; use fable-experimenter for original build prompts.
Subagents overview
What people ask about claude-directory
What is pulkitxm/claude-directory?
+
pulkitxm/claude-directory is subagents for the Claude AI ecosystem. Open-source AI interfaces built with Claude (Fable 5) including hero sections, GLSL shaders, design systems, animations, 3D components and and landing pages, in React, Tailwind, and Three.js. Plug them into your AI agent and ship faster. It has 539 GitHub stars and its last recorded update is dated 2026-09-11.
How do I install claude-directory?
+
You can install claude-directory by cloning the repository (https://github.com/pulkitxm/claude-directory) or following the README instructions on GitHub. ClaudeWave also provides quick install blocks on this page.
Is pulkitxm/claude-directory safe to use?
+
Our security agent has analyzed pulkitxm/claude-directory and assigned a Trust Score of 97/100 (tier: Verified). See the full breakdown of passed checks and flags on this page.
Who maintains pulkitxm/claude-directory?
+
pulkitxm/claude-directory is maintained by pulkitxm. The last recorded GitHub activity is dated 2026-09-11, with 27 open issues.
Are there alternatives to claude-directory?
+
Yes. On ClaudeWave you can browse similar subagents at /categories/agents, sorted by popularity or recent activity.
Deploy claude-directory to your cloud
Ship this repo to production in minutes. Each platform spins up its own environment with editable env vars.
Maintain this repo? Add a badge to your README
Drop the badge into your GitHub README to show it's tracked on ClaudeWave. Each badge links back to this page and reflects the live Trust Score.
[](https://claudewave.com/repo/pulkitxm-claude-directory)<a href="https://claudewave.com/repo/pulkitxm-claude-directory"><img src="https://claudewave.com/api/badge/pulkitxm-claude-directory" alt="Featured on ClaudeWave: pulkitxm/claude-directory" width="320" height="64" /></a>More Subagents
The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
The agent that grows with you
Java 面试 & 后端通用面试指南,覆盖计算机基础、数据库、分布式、高并发、系统设计与 AI 应用开发
Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or self-hosted, so teams move from prototype to production without rebuilding the stack.
The agent engineering platform.
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.