Shape agent capabilities into your workflow.

AgentBox moves a project and supported agent settings into isolated environments so Claude Code, Codex, OpenCode, and similar tools can run in parallel on local Docker or supported cloud providers. Each box offers persistent shells plus browser and editor access, can resume quickly from checkpoints, and keeps Git credentials on the host. Local use requires macOS or Linux, Docker, and Node 20.10 or newer.

public-agent-skills shares one author's working collection for development, documentation, content, research, and automation tasks, plus a script that copies skills into several agent-specific directories. It is useful as a reference for practical personal workflows rather than a universal catalog. The repository explicitly recommends reviewing every document, script, and dependency before installation because the skills are highly individualized.

Hermes Agent Idea Workflow structures the pre-build phase before any coding begins. It can capture an idea in a lightweight mode or guide a full interview that produces a design document, an optional UI brief, an implementation specification, and a single-file handoff for another agent. A review gate marks the package as pass, pass with changes, or fail before handing it to a Superpowers-based build process. The workflow itself does not implement the product and remains in beta field testing.

Skyll exposes a REST API and MCP server that searches multiple skill sources, ranks matches, and returns complete SKILL.md content with optional references. Agents can choose and load instructions during a task instead of relying on preinstalled skills. A hosted service is available, or teams can self-host the server for more control.

Patterns.dev Skills packages 58 development patterns for JavaScript, React, and Vue as independently installable agent skills. Users can add an entire technology stack or select individual topics such as hooks, singletons, composables, AI interface patterns, streaming SSR, code splitting, or rendering strategies. The collection gives coding agents concise prose and code examples for implementation decisions. It deliberately omits the animations, diagrams, sandboxes, and other visual walkthroughs available on the full patterns.dev website.

humanize contains two static, dependency-free agent skills. One rewrites or generates prose using nine research-derived levers for rhythm, specificity, voice, and structure; the other examines nine categories of signals associated with AI-generated writing. The rules draw on more than 50 sources and work across many agents, though any detection-style judgment remains sensitive to changing methods and imperfect signals.

This collection gives Claude 31 reusable workflows for analytics work, from validating data and SQL to diagnosing metric changes and communicating results. After a user describes the task, Claude selects an appropriate workflow, asks for the minimum context, states assumptions, and returns a consistent output; when inputs are incomplete, it proceeds with those gaps made explicit.

paper2anything packages five independent skills for turning an academic PDF into a PowerPoint deck, conference poster, single-page project site, Xiaohongshu post, or WeChat article. It is useful when a research team needs several dissemination formats from the same paper without rebuilding each workflow separately. Paper extraction uses MinerU credentials, and some rendering or visual-check paths also require local tools such as LibreOffice, Poppler, and Node.js.

Micro-Drama-Skills provides a Claude Skills workflow for producing a 25-episode short drama, with roughly 30 seconds per episode. It creates scripts, character definitions, six-panel storyboards, and project configuration, then can call Gemini and Veo for images and video before submitting batches to a Seedance workflow. The project is intended for tools that load Claude Skills and requires users to configure the Python environment and the relevant media-generation APIs.

SkillCompass is a local-first evaluator and manager for people who maintain agent skills. It scores a skill across six dimensions, identifies the weakest area to improve, and tracks versions so later evaluations show whether a change helped. Usage hooks feed an inbox that surfaces unused skills, stale evaluations, declining adoption, and updates, while SKILL.md edits trigger structural and security-pattern checks. SHA-256 snapshots make regressions recoverable. It is not a general code reviewer, runtime debugger, or tool for creating skills from scratch.

Awesome Apify Skills is a community collection of SKILL.md packages for AI coding assistants. Each skill teaches an agent which Apify Actors to use, how to prepare inputs, and how to handle failures for a specific web-data or research task, including areas such as advertising intelligence and search visibility. Users can install the whole repository or select one skill, and the packages work across Claude Code, Codex, Cursor, Gemini CLI, and many other compatible agents. Running the workflows generally requires an Apify account and authentication token.

Blueprint packages a focused software-delivery process into agent skills for design documents, specifications, planning, implementation, testing, browser verification, and code review. Teams can invoke one stage or coordinate tickets through review-ready pull requests. Completion is tied to tests and observable proof, and browser-based checks require a compatible automation setup.

GRACE Marketplace distributes a contract-first set of AI engineering skills plus an optional CLI; it is not a standalone end-user application. The workflow anchors planning, implementation, verification, review, and refactoring in shared XML artifacts for requirements, technology choices, development plans, verification evidence, and a knowledge graph. The skills target Claude Code, Codex CLI, and Kilo Code, while the CLI mainly adds integrity checks and module navigation around the same artifacts.

Claude Trading Skills is a collection of 67 Agent Skills for market data, on-chain research, technical analysis, backtesting, portfolio and risk work, and other quantitative-finance workflows. The library is crypto and DeFi first, while its structure can be extended to additional markets. It requires Python, uv, and an Agent Skills-compatible client, with API credentials needed for some data providers. Skills that can interact with live markets default to demonstration or simulation behavior and require explicit confirmation before execution.

Bitget Wallet Skill gives agents tools for token discovery, market and address analysis, balance queries, swaps, transfers, and order tracking across supported EVM chains and Solana. Fund-moving workflows generate transaction data and require user confirmation and signing by default. Using it depends on Bitget Wallet's API, network access, and appropriate wallet credentials.

SuperSEO Skills lets an agent start with a URL, keyword, or topic, research the page, search results, and competitors, then produce SEO audits, content plans, drafts, and other growth work. It is built for live, one-off research workflows rather than historical Search Console analysis or ongoing site monitoring.

This repository provides a lightweight meta-skill rather than duplicating Streamlit's full development guidance. It detects the active Python interpreter, locates the installed Streamlit package, and points an AI assistant to bundled skills for dashboards, themes, layouts, session state, and components. Streamlit 1.57 and later ship those skills inside the pip package, while 1.58 adds the `streamlit skills` installer for project or global setup. Older versions fall back to online documentation, so the local workflow depends on an installed Streamlit environment.

NextClaw is a local-first AI workspace for tasks that span files, models, agents, browser tools, local applications, and messaging channels. It keeps references, generated results, and follow-up work together, while separate agents can have their own roles, memory, skills, runtimes, and workspaces. It can support research, data analysis, document drafting, file processing, and scheduled jobs through desktop or Docker deployments. Local configuration and credentials stay in the controlled environment, although connected providers receive the data sent to them.