Compare 44 workflows by task, requirements and tradeoffs before giving an agent access to your files or accounts.
Each profile records a source commit and review scope. We have not independently installed or executed these skills. Source checks are not security certification.
Browse all 44 profiles Submit a skillProductivity · Codex source
A document workflow for reading, generating and visually reviewing PDF files. It combines text extraction with page rendering so layout problems are checked before delivery.
Source checked 2026-10-20 · Not runtime tested
Design · Claude Code source
Helps an agent choose a deliberate visual direction when building or reshaping a frontend. The instructions emphasize grounding typography, layout and styling in the real product brief, then reviewing the result for generic or mismatched choices.
Source checked 2026-10-20 · Not runtime tested
SEO & marketing · Format-based source
Organizes an SEO diagnosis around crawlability, technical foundations, page content and evidence from the site’s current performance. It can turn an unclear traffic problem into a prioritized investigation rather than a generic list of keyword changes.
Source checked 2026-10-20 · Not runtime tested
Development · Format-based source
Organizes React and Next.js performance guidance by impact, including request waterfalls, bundle size, server work and rendering. It gives an agent concrete review references while leaving actual performance gains to measurement in the target application.
Source checked 2026-10-20 · Not runtime tested
Data & research · Codex source
Searches and compares authorized Notion pages, then turns findings into a brief or report with links back to the source material. It combines research structure with a connected workspace rather than acting as an unrestricted web researcher.
Source checked 2026-10-20 · Not runtime tested
Productivity · Claude Code source
Selects a communication format and loads the matching example guidance for updates, newsletters, FAQs or other internal messages. It provides a repeatable starting structure that still needs the organization’s facts and voice.
Source checked 2026-10-20 · Not runtime tested
Editorial guide · 2026-10-20
A practical selection method for agent skills: define the output, inspect dependencies and permissions, compare alternatives, and run a small evidence-based trial.
Editorial guide · 2026-10-20
Install a complete skill folder in the right scope, confirm discovery separately from dependencies, and troubleshoot Codex, Claude Code and Cursor without guessing compatibility.
Editorial guide · 2026-10-20
Understand the practical boundary between agent skills and MCP, with examples for Figma, Notion, CI diagnosis and local document tasks, plus a dependency worksheet.
Editorial guide · 2026-10-20
Build a controlled skill trial with synthetic fixtures, observable acceptance checks, a no-skill baseline and a downloadable results ledger. Distinguish discovery from verified task quality.
| You need | Look for | Check before accepting |
|---|---|---|
| A document or report | Source handling, output format and rendering steps | Accurate citations, complete sections and a readable final file |
| A code change | Repository context and focused development guidance | A minimal diff, relevant checks and no unrelated edits |
| A marketing asset | An approved brief and an explicit audience | Supported claims, correct offer and a usable channel format |
| A browser or connected task | Required tools, accounts and action boundaries | The resulting system state and any external action |
A skill provides reusable workflow instructions and may include supporting resources. A tool performs an operation. An MCP connection can expose tools or information to a compatible host. Installing instructions does not automatically supply a browser, API key or paid account. Check these dependencies before comparing two skills as substitutes.
Use a non-sensitive sample with a known result. Ask for one deliverable, inspect it and request a revision. Record manual corrections and access requested. If the workflow cannot explain a missing dependency or distinguish a failed action from a completed one, resolve that gap before expanding to real work. The profile checklists help you evaluate the artifact rather than the confidence of the completion message.
We record a specific source commit, read its workflow and declared license evidence, and distinguish author claims from independent testing. Each profile includes a proposed trial and acceptance checks. Unknown terms, moving references and untested compatibility stay visible.
The collection draws from OpenAI, Anthropic, Vercel and Corey Haines repositories. A publisher's identity does not guarantee the result of a task or establish that every referenced dependency was reviewed.
Choose a workflow with a checkable output, use synthetic input and review its access needs. Follow the installation guide for your host, then compare the actual artifact with an unchanged baseline. The downloadable worksheets help record both success and unresolved checks.
Submit a listing for editorial review. Include a live website, current pricing and a clear use case. We may request more information before publication.
