floatboat
Open navigation menu

Skills Marketplace

Find software projects and Skills to help design your own agent workflows. Explore task structure, tool connections, context, and result review through representative work from organizations and creators, then plan a focused experiment around your work in Floatboat.

38 Projects30 Skills63 Publishers

Projects and Skills are references for workflow design. Validate any integration, runtime, and compatibility requirements against your environment and the upstream documentation.

Teams and Creators Behind the Work

Meet the organizations and creators building coding agents, runtimes, creative tools, and Skills. Explore their focus, representative projects, and maintained sources before choosing resources for development, design, video, or office work.

Jump to projects and Skills ↓

46 results

  • Vercel Labs

    Vercel Labs explores developer tools and public experiments, including Agent Skills, browser automation, and generative interfaces for building and testing software.

    1 projects · 0 Skills
  • Anthropic

    Anthropic develops Claude and publishes developer resources, including Claude Code, application examples, and Skills for design, documents, authoring, and testing workflows.

    0 projects · 7 Skills
  • HeyGen

    HeyGen builds AI video products and publishes tools including HyperFrames and a video API CLI, connecting visual composition, media preparation, and rendering workflows.

    1 projects · 1 Skills
  • Microsoft

    Microsoft publishes development tools, languages, and cloud resources, including Azure Skills, Visual Studio Code, and TypeScript for application development, deployment, and diagnostics.

    1 projects · 0 Skills
  • Cloudflare

    Cloudflare builds network infrastructure and developer tools, with public projects spanning the Workers runtime, protocol implementations, and Agent instructions for structured code security audits.

    0 projects · 1 Skills
  • Firecrawl

    Firecrawl develops web data tools for agents, connecting search, scraping, crawling, and structured extraction through its main platform, CLI, and task-focused instructions.

    1 projects · 0 Skills
  • Supabase

    Supabase builds a PostgreSQL application platform and publishes Agent Skills for database design, query performance, security, and practical integration with backend services and tools.

    1 projects · 0 Skills
  • OpenAI

    OpenAI develops AI systems and developer tools. Explore Codex, its open agent harness for coding tasks, tool use, and application integration.

    1 projects · 0 Skills
  • DeepSeek

    DeepSeek publishes agent infrastructure and research prototypes. Explore Harness for runtime composition and DSH Superagent for GitHub-driven coding workflows.

    1 projects · 0 Skills
  • Earendil Works

    Earendil Works hosts Pi’s current upstream. Explore its extensible agent harness, coding CLI, and reusable libraries for model access and agent execution.

    1 projects · 0 Skills
  • Anomaly

    Anomaly publishes OpenCode, an open-source coding agent. Explore its interactive developer workflow, official source, and the boundary between application tools and model services.

    1 projects · 0 Skills
  • Nous Research

    Nous Research publishes Hermes Agent. Explore a system combining persistent memory, reusable Skills, connected tools, and scheduling for work that continues across sessions.

    1 projects · 0 Skills
  • Comfy Org

    Develops ComfyUI and its open ecosystem for node-based generative-media workflows.

    1 projects · 0 Skills
  • Graphify Labs

    Develops Graphify’s code knowledge graphs, local tooling, and agent integrations.

    1 projects · 0 Skills
  • Stably

    Publishes Orca, a workspace for parallel coding agents, isolated worktrees, and code review.

    1 projects · 0 Skills
  • OpenDesign

    Publishes an open-source design workspace and reusable packages for agent-assisted design.

    1 projects · 0 Skills
  • Sherlock Project

    Maintains Sherlock, a command-line tool for checking public username matches across supported websites and exporting candidates for review.

    1 projects · 0 Skills
  • K-Dense AI

    Maintains Scientific Agent Skills for data inspection, statistical analysis, visualization, and scientific writing in supported Agent hosts.

    1 projects · 1 Skills
  • OpenClaw Foundation

    Maintains the OpenClaw agent runtime, Gateway, and headless CLI for tasks with explicit workspace and tool access.

    1 projects · 0 Skills
  • Skyvern

    Builds Skyvern’s browser workflow platform and coding-agent interfaces, with cloud and self-hosted execution options.

    1 projects · 0 Skills
  • Steel

    Publishes Steel Browser infrastructure and official cloud-browser Skills for live browsing, reusable code, and session diagnosis.

    1 projects · 0 Skills
  • Bright Data

    Publishes agent connectors and official Skills for web access, search, extraction, and browser services backed by Bright Data.

    1 projects · 0 Skills
  • Lightpanda

    Maintains the Lightpanda headless browser and a separate official agent Skill, with MCP access and platform-specific guidance.

    1 projects · 0 Skills
  • CitroLabs

    Maintains Ego Lite and ego-browser tooling for sharing browser work with external coding agents, with task Spaces and login-state access.

    1 projects · 0 Skills
  • BrowserOS

    Publishes BrowserOS and neo’s external-agent MCP connection, with browser-specific Skills and client setup documentation.

    1 projects · 0 Skills
  • Browser MCP

    Maintains Browser MCP’s local server and Chrome extension, connecting coding agents to approved existing browser profiles.

    1 projects · 0 Skills
  • PinchTab

    Maintains a browser control server with CLI, HTTP, MCP, and official Skills, focusing on profiles, instances, and task sessions.

    1 projects · 0 Skills
  • Browser Use

    Maintains Browser Use and its coding-agent tooling, with local browser control, a Python framework, and hosted browser services.

    1 projects · 0 Skills
  • Browserbase

    Builds Stagehand and browser infrastructure, connecting deterministic code, agent interaction, and local or hosted browser workflows.

    1 projects · 0 Skills
  • Chrome DevTools

    Publishes Chrome DevTools MCP and official debugging Skills for browser inspection, runtime evidence, and coding-agent workflows.

    1 projects · 0 Skills
  • Adobe

    Adobe publishes creative-tool Skills and React Spectrum libraries, connecting product-specific agent guidance with accessible interface development.

    Organization profile
  • Amazon Web Services

    AWS maintains an agent toolkit combining cloud-development instructions, MCP servers and plugins for working with AWS services.

    Organization profile
  • Google Gemini

    The google-gemini organization publishes Gemini API Skills and Gemini CLI, separating reusable API guidance from the terminal agent.

    Organization profile
  • Google Labs Code

    Google Labs Code maintains Stitch design Skills and plugins, with host-specific setup documented in its own repositories.

    Organization profile
  • Firebase

    Firebase publishes product-specific agent instructions for building applications, with official documentation linking to its agent-skills repository.

    Organization profile
  • Hugging Face

    Hugging Face publishes AI and ML task Skills covering dataset creation, model training, evaluation and Hub operations.

    Organization profile
  • MongoDB

    MongoDB maintains database software and official agent Skills, with separate Atlas and self-managed connection paths.

    Organization profile
  • Redis

    Redis publishes server software and agent instructions for data structures, key design, memory, expiration and related application decisions.

    Organization profile
  • Stripe

    Stripe publishes agent plugins and Skills alongside payment SDKs; integration instructions, API clients and hosted services have distinct roles.

    Organization profile
  • Shopify

    Shopify’s AI Toolkit connects agents to developer documentation, API schemas and code validation, alongside separate Shopify CLI tooling.

    Organization profile
  • Figma

    Figma’s MCP guide documents design context and agent workflows, with access conditions and developer terms distinct from source-code availability.

    Organization profile
  • Notion

    Notion publishes Skills and its JavaScript API client under makenotion, separating agent instructions from application integration code.

    Organization profile
  • Datadog

    The DataDog organization maintains Datadog Agent software; experimental AI Skills belong to the separate Datadog Labs organization.

    Organization profile
  • Datadog Labs

    Datadog Labs publishes experimental AI Skills under its own GitHub organization, with an explicit boundary from Datadog product support.

    Organization profile
  • Sentry

    Sentry separates employee development Skills from product-focused agent plugin sources, all under the getsentry GitHub organization.

    Organization profile
  • HashiCorp

    HashiCorp publishes Terraform and Packer agent instructions alongside infrastructure software, with licenses and execution requirements checked per repository.

    Organization profile

Projects and Skills

Compare coding agents, design tools, and media software by task, or switch to Skills for reusable instructions. Search by project, Skill, or publisher if you know the name.

Browse task categories →

38 results

  • Impeccable

    Paul Bakaus

    Install design guidance across supported coding agents and run the supporting engine for project context, interface checks, and browser-based design iteration.

    Local guide
  • Superpowers

    Jesse Vincent

    Connect composable engineering Skills with a software development workflow and integrations for coding agents, covering design, implementation, review, and verification.

    Local guide
  • agent-browser

    Vercel Labs

    Open a page, read a compact accessibility snapshot, and act on element references from the terminal. agent-browser keeps the browser session available across commands.

    Local guide
  • REA

    morluto

    Use agents to investigate applications and native binaries, connecting observed behavior to evidence about how the software works.

    Local guide
  • HyperFrames

    HeyGen

    Build video scenes with HTML and agent-oriented tooling, connecting visual composition and timed rendering in a programmable workflow.

    Local guide
  • Caveman Engine

    Julius Brussee

    Explore compression software within the Caveman project, assessing information recovery alongside output size for downstream processing tasks.

    Local guide
  • Playwright MCP

    Microsoft

    Connect a coding agent to Playwright through MCP for navigation, accessibility snapshots, and browser interactions. Compare the persistent MCP workflow with the separate Playwright CLI and its official Skill.

    Local guide
  • Firecrawl

    Firecrawl

    Search, scrape, and interact with web content through a platform that returns Markdown, structured data, and other outputs for applications and agents.

    Local guide
  • Supabase

    Supabase

    Build applications on a Postgres-backed platform with authentication, storage, APIs, functions, and realtime capabilities, using hosted or self-hosted deployment workflows.

    Local guide
  • Codex

    OpenAI

    Study Codex’s context, tool loop, and approval model to improve your own agent workflows, with CLI, SDK, and app-server options when runtime integration is needed.

    Local guide
  • DeepSeek Harness

    DeepSeek

    Study DeepSeek Harness’s plugin architecture to compose task capabilities, connect tools, and define clear handoffs when designing or extending your own agent workflows.

    Local guide
  • Pi

    Earendil Works

    Use Pi’s minimal core, Skills, extensions, and runtime libraries as references for choosing the right level of customization in your own agent workflows.

    Local guide
  • OpenCode

    Anomaly

    Study OpenCode’s task modes, tool feedback, and review flow to improve repository investigation, implementation, and validation stages in your own development workflows.

    Local guide
  • Hermes Agent

    Nous Research

    Study Hermes Agent’s memory, Skills, tools, and scheduling to improve context continuity, reusable procedures, and recurring tasks in your own workflows.

    Local guide
  • Ponytail

    DietrichGebert

    Reuse-first coding instructions, change review, and repository audits for existing agents, with host-specific plugins and portable rules.

    Local guide
  • ComfyUI

    Comfy Org

    Build and reuse node-based generative-media workflows, preserving models, custom-node dependencies, and reviewable outputs.

    Local guide
  • Graphify

    Graphify Labs

    Build queryable code knowledge graphs with a local CLI and agent guidance, keeping relationship provenance and source revision visible.

    Local guide
  • Orca

    Stably

    Coordinate coding agents in separate worktrees, compare their changes, and feed review findings back into bounded tasks.

    Local guide
  • OpenMontage

    Calesthio

    Plan, assemble, and review videos with a coding agent, from footage selection to rendered checks.

    Local guide
  • OpenDesign

    OpenDesign

    Create and review prototypes, decks, and motion with an agent and a reusable design system.

    Local guide
  • Claude-Mem

    Alex Newman

    Summarize coding-session observations and retrieve selected history through local search or MCP.

    Local guide
  • Autoresearch

    Andrej Karpathy

    Run single-GPU model experiments with a coding agent, then compare changes under a fixed training budget.

    Local guide
  • Sherlock

    Sherlock Project

    Check approved usernames across supported sites, export public candidate links, and review them against account records.

    Local guide
  • Scientific Agent Skills

    K-Dense AI

    Install research Skills for data inspection, statistical analysis, visualization, and writing in Codex, Cursor, and other supported hosts.

    Local guide
  • OpenClaw

    OpenClaw Foundation

    Run bounded agent tasks through a headless CLI or a persistent Gateway session.

    Local guide
  • ECC

    Affaan Mustafa

    Preview coding-agent setup, inspect host-specific hooks, and check installed configuration.

    Local guide
  • Skyvern

    Skyvern

    Use Skyvern’s MCP or API for browser operations and multi-step workflows. Choose cloud or self-hosted execution, supply the required credentials, and review the business result rather than only the run status.

    Local guide
  • Steel Browser

    Steel

    Use Steel to provide a browser session for agent work. Compare the open-source browser runtime with Steel’s hosted MCP and official Skills, keeping service authentication and session costs explicit.

    Local guide
  • Bright Data MCP

    Bright Data

    Connect a coding agent to Bright Data’s web access and data tools. Use its official Skills to choose browser, fetch, search, or extraction workflows, with credentials, product scope, and costs checked separately.

    Local guide
  • Lightpanda

    Lightpanda

    Use Lightpanda for headless page access, extraction, and supported interactions through native MCP or official agent guidance. Check compatibility before replacing a full browser, especially for visual or layout-dependent work.

    Local guide
  • Crawl4AI

    UncleCode

    Collect page content for research or data processing through Crawl4AI’s Python, deployed API, or documented MCP interface. Define the crawl boundary and validate extracted fields before using the data.

    Local guide
  • Ego Lite

    CitroLabs

    Use the ego-browser Skill to let a coding agent work in its own browser Space while you keep your tabs. Ego Lite supports macOS and Windows and can reuse imported login state.

    Local guide
  • BrowserOS

    BrowserOS

    Use BrowserOS neo’s local MCP endpoint from Floatboat or another MCP host. Keep the browser product, the external-agent connection, and the official browseros-neo Skill in one workflow.

    Local guide
  • Browser MCP

    Browser MCP

    Connect a local MCP server and Chrome extension so an agent can use your existing browser profile. Pair the extension deliberately, then work in task tabs with the selected account.

    Local guide
  • PinchTab

    PinchTab

    Run a local browser control plane and call it through CLI, HTTP, or MCP. PinchTab separates profiles, browser instances, and task tabs, with an official Skill for coding agents.

    Local guide
  • Browser Use

    Browser Use

    Give an agent a local or cloud browser for multi-step web tasks. Choose the coding-agent CLI or Python framework, then inspect the resulting page and output.

    Local guide
  • Stagehand

    Browserbase

    Use Stagehand when a browser workflow needs both explicit code and language-driven interaction. Coding agents can start with the browse CLI and official Skill; reusable applications can use the SDK.

    Local guide
  • Chrome DevTools MCP

    Chrome DevTools

    Let a coding agent inspect live Chrome with console, network, performance, and page tools. Use the evidence to diagnose a particular failure rather than treating a screenshot as the whole investigation.

    Local guide

From Skill Instructions to Finished Work

A useful Skill still needs project material, tools, and a result someone can review. These Floatboat capabilities connect that method with a persistent workspace, making the task context and delivered work available for the next revision.

Agent Workspace

Keep the task brief, selected Skill instructions, and project materials together. Define the inputs, steps, and expected result so research, implementation, and review continue within the same project.

Follow-up revisions can use existing files, discussions, and outputs without repeatedly rebuilding the project brief. Third-party tools, runtime dependencies, and available capabilities still need to be configured and checked against their own documentation.

FloatboatProject execution and review

Project files

▱ Project brief.md

▱ References/

▱ Task instructions.md

Agent run

Implementation planEditable draftReview notes

Flow Mode

Discuss the brief while revising a plan, page copy, or draft deliverable. Adapt general task instructions to the needs of the project and bring feedback into the work as it develops.

Moving from a method to a finished result often involves clarifying goals and refining details. Ongoing conversation keeps revisions connected to the existing work before you check whether the outcome meets the brief.

FloatboatProject execution and review

Listening and editing

Implementation plan

AI File Organizer

Organize local reference documents, project assets, and intermediate outputs with clear folders and filenames. Make task inputs easier to find and keep materials understandable for later review and handoff.

Choose the folders in scope, then inspect the proposed naming and grouping. Recognizable sources and versions help distinguish reference materials, working files, and final deliverables as the project develops.

FloatboatProject execution and review

Before

□ 1_Project brief.md_final

□ 2_References/_final

□ 3_Task instructions.md_final

Suggested structure

└ Project brief.md

└ References/

└ Task instructions.md

Multiple AI Models

Choose models for reading sources, comparing plans, drafting content, or reviewing results while retaining the same project files, requirements, and revision history throughout the work.

Start with what the current step needs to understand or produce, then choose a model for that task. Model selection, the Skill’s tool requirements, and verification of the final deliverable remain separate decisions.

FloatboatProject execution and review

Auto

Automatic routing

Same workspace context

Claude Opus 4.6

GPT-6 Astra

Gemini 3.1 Pro

Frequently Asked Questions

What is the difference between a publisher, a project, and a Skill?
A publisher is the person or organization responsible for the work. A project may be a CLI, library, framework, application, or other resource. A Skill supplies task instructions for an agent. Use the publisher section to understand the source, then inspect the specific item.
How should I start when I know the task but not the tool name?
Choose Projects or Skills, then narrow the directory by task category and tags such as React, code review, or browser automation. Category pages show instructions and independent projects in separate sections so you can compare their roles.
Where do the directory links go?
Every listed Skill opens a local guide. Projects open a local detail page when available or link to the original source. The last column identifies the destination type.
Does open-source software include free models and services?
No. Source licensing is separate from model APIs, subscriptions, cloud services, and third-party components. An open agent or harness does not include free model credits, an installed runtime, or permission to access external systems.
Where can I check who maintains an item?
Open its publisher profile and original repository or website. Review ownership, documentation, releases, and current issues there. The publisher profile identifies the source of the work, while a project’s current maintenance status must be checked upstream.
Does the directory order represent popularity or verified quality?
No. Independent projects with local detail pages appear first to support workflow comparison; that is not a popularity or quality ranking. Repository stars, usage, and installation counts measure different things. Choose by source, maintenance, and task requirements.

Give Your Workflow a Workspace

See how Floatboat brings agents, local files, tools, and deliverables into one workspace.

Meet the Agent Workspace