Top 10 Codex Skills in 2026: Weekly Rankings and Builder Picks

The most useful Codex skills are reusable workflows that make repeated work more reliable, but “best” and “trending” are not the same thing. This guide combines a stable editor-recommended stack for software builders with a July 24, 2026 snapshot of the skills gaining short-term installation momentum. Use the trend list to discover new capabilities, then choose what to install based on your actual workflow, source quality, permissions, and maintenance status.

If you want to compare skills by platform and use case, the AI Agent Skill Finder can help you narrow the list before installing anything.

What Makes a Skill Good for Codex?

A good Codex skill should solve a repeatable workflow. It needs a clear trigger, a narrow purpose, and a reason to exist outside a normal prompt. If you repeat the same process every week—such as resolving merge conflicts, testing browser flows, reviewing security risks, or producing videos—it may deserve a dedicated skill.

Codex supports skills as reusable packages of instructions, resources, assets, and optional scripts. A skill normally includes a SKILL.md file that explains when it should activate and how the workflow should run. The official Codex skills documentation explains the underlying structure.

Codex Skills vs Codex Plugins

A skill defines a reusable workflow. A plugin is a broader installable package that can include skills alongside commands, tools, integrations, apps, configuration, and other supporting resources.

A local skill is usually enough for a personal workflow. A plugin becomes more useful when a team needs to distribute the same workflow together with approved integrations, tools, or workspace controls.

Codex Skills vs AGENTS.md

AGENTS.md is best for durable repository instructions such as project structure, test commands, coding conventions, and branch rules. Skills are better for reusable workflows that may travel across repositories.

For example, an AGENTS.md file might instruct Codex to run a particular test command before completing changes. A skill can define the complete process for investigating a failed CI check, collecting evidence, proposing a fix, applying an approved change, and verifying the result.

Popularity vs Practical Value

A trending leaderboard measures short-term attention or installation activity. It does not prove that a skill is secure, actively maintained, compatible with Codex, or appropriate for every project. A specialized media-generation skill may trend quickly while a less visible CI or security skill creates more lasting value for a software team.

For that reason, the two lists below serve different purposes. The weekly snapshot helps you discover what builders are exploring, while the editor-recommended list focuses on repeatable development workflows.

Codex Skills Trending in the Week of July 24, 2026

The following snapshot is based on the July 24, 2026 weekly ranking shown in the supplied reference image, which identifies skills.sh trending data as its source. Rankings can change rapidly and several entries are third-party projects rather than OpenAI-maintained skills.

Rank	Skill	Primary workflow	Consider it when
1	ai-video-generation	Coordinates models and tools for AI video generation	You repeatedly create videos and need a reusable production workflow
2	find-skills	Searches for skills that match a requested task	You need discovery before deciding what to install
3	design-guide	Generates and checks consistent interface guidelines	You want repeatable visual rules across generated interfaces
4	grill-me	Challenges an idea through sustained questioning	You want to pressure-test assumptions before implementation
5	prisma-client-api	Supports Prisma queries and data operations	Your application uses Prisma and repeated database patterns
6	resolving-merge-conflicts	Analyzes and resolves Git merge conflicts	You need a structured conflict-resolution workflow
7	reddit-automation	Searches, reads, and organizes Reddit content	Your research workflow depends on Reddit discussions
8	hyperframes-cli	Creates, previews, and renders programmatic videos	You want to produce video through an agent-driven HTML workflow
9	ai-music	Generates songs, scores, or reusable audio assets	Your workflow repeatedly requires AI-generated music
10	agent-browser	Controls a browser to complete webpage tasks	Your agent must navigate, inspect, or interact with websites

Why ai-video-generation Ranked First

ai-video-generation addresses a complete creative workflow rather than a single coding operation. Its appeal comes from coordinating the steps and tools involved in generating AI video. That can be valuable for builders producing explainers, product demos, social content, or visual prototypes repeatedly.

Its ranking does not mean every developer should install it first. Video workflows may involve external services, media uploads, API credentials, generation costs, and licensing questions. Inspect the skill before giving it access to production assets or paid accounts.

Why find-skills Is a Practical Starting Point

find-skills focuses on discovery. Instead of assuming which skill fits a task, it helps search for relevant options. This makes it useful for builders entering a large skills ecosystem without knowing the exact project or repository name they need.

Discovery is only the first step. A result still needs to be checked for source ownership, recent activity, installation commands, scripts, permissions, and compatibility with the agent environment being used.

Why design-guide Is Useful for Interface Work

design-guide targets a common weakness in AI-generated interfaces: individual pages may look acceptable while failing to follow a consistent visual system. A reusable design workflow can give an agent clearer rules for typography, spacing, components, colors, responsive behavior, and interface review.

This type of skill is most useful when it reflects a real project design system. Generic visual preferences should not override existing brand tokens, accessibility requirements, or repository-specific instructions.

Specialized Skills Need Specialized Evaluation

The rest of the weekly list covers very different jobs. grill-me helps interrogate an idea, prisma-client-api focuses on database work, and resolving-merge-conflicts supports a constrained Git task. Media-oriented entries such as hyperframes-cli and ai-music solve creative production problems, while agent-browser introduces browser-control capabilities.

These skills should not be compared on one universal quality scale. The correct question is whether a skill reliably completes a workflow you actually repeat and whether the access it requests is proportionate to that task.

Top 10 Codex Skills for a Stable Builder Workflow

The following editor-recommended list is designed for longer-term software-building workflows. It prioritizes focused tasks, verifiable outcomes, and sources that can be inspected rather than short-term popularity.

1. skill-creator

skill-creator helps turn a successful one-off session into a reusable workflow. It can preserve review rules, test commands, release procedures, writing examples, and validation steps that would otherwise need to be pasted into every new task.

Best for: creating custom Codex workflows and capturing repeated procedures.

Why it matters: most teams do not need hundreds of unrelated public skills. They need a small set that accurately represents the work they perform repeatedly.

Install first if: you keep reusing the same review checklist, release process, testing procedure, or content workflow.

2. gh-fix-ci

gh-fix-ci is designed to investigate failing GitHub pull-request checks. It can inspect checks and logs, summarize the failure, prepare a fix plan, apply an approved change, and recheck the result.

Best for: GitHub Actions failures, PR checks, CI debugging, and test-log triage.

Why it matters: it gives Codex an evidence-driven loop instead of encouraging it to guess from a partial error message.

Install first if: your team regularly loses time to lint, test, build, or type-check failures.

3. gh-address-comments

gh-address-comments helps Codex review comments on the open pull request for the current branch, summarize the requested changes, and address selected feedback.

Best for: PR review cleanup, reviewer feedback, issue comments, and focused GitHub collaboration.

Why it matters: addressing a review comment is a bounded task. The agent needs to understand the exact thread and change only the relevant code.

Install first if: you use Codex to update pull requests after human review.

4. playwright-interactive

playwright-interactive supports persistent browser sessions for interactive web and Electron application testing.

Best for: web-app QA, Electron testing, UI debugging, viewport checks, and visual verification.

Why it matters: compiling successfully does not prove that an interface works. A browser-testing workflow lets Codex click controls, inspect application states, evaluate layout behavior, and collect evidence.

Install first if: you build interfaces, dashboards, internal tools, or web products.

5. security-best-practices

security-best-practices provides a structured workflow for language- and framework-specific security reviews.

Best for: secure-by-default coding, vulnerability reports, framework review, and security triage.

Why it matters: functional code is not automatically safe. Security review requires different evidence, priorities, and validation than a normal code-quality pass.

Install first if: Codex works on authentication, permissions, tokens, uploads, webhooks, public identifiers, or backend APIs.

6. sentry

sentry supports read-oriented production issue investigation using Sentry data.

Best for: unresolved issue review, event inspection, error summaries, and incident triage.

Why it matters: production debugging should begin with real event context instead of a vague description of what might have failed.

Install first if: your application already uses Sentry and you want Codex to investigate evidence before proposing code changes.

7. Vercel Agent Skills

Vercel Agent Skills cover workflows associated with React, Next.js, frontend quality, performance, and Vercel-hosted applications.

Best for: React applications, Next.js, accessibility, performance review, and deployment optimization.

Why it matters: generated frontend code may work while still producing unnecessary waterfalls, large bundles, weak accessibility, or inefficient server behavior.

Install first if: you build SaaS frontends, dashboards, marketing sites, or applications deployed through Vercel.

8. Supabase Agent Skills

Supabase Agent Skills support workflows involving Postgres, authentication, row-level security, migrations, storage, vectors, functions, and database performance.

Best for: SaaS backends, database design, RLS policies, migrations, indexes, and Supabase development.

Why it matters: AI-generated backend code commonly fails around permissions, schema design, and unsafe assumptions about data access.

Install first if: your application uses Supabase for authentication, storage, database, or backend services.

9. Cloudflare Skills

Cloudflare Skills support development across Workers, Pages, Durable Objects, Wrangler, R2, D1, KV, Vectorize, tunnels, and related edge workflows.

Best for: edge applications, Workers, APIs, agents, storage bindings, and Cloudflare deployment.

Why it matters: platform-specific skills can prevent Codex from generating generic serverless code that ignores the actual bindings, storage model, caching behavior, or deployment configuration.

Install first if: you deploy agents, APIs, dashboards, or MCP services through Cloudflare.

10. n8n Official Skills

n8n skills support workflow design, expressions, loops, pagination, sub-workflows, AI agent nodes, error handling, credentials, and automation debugging.

Best for: self-hosted automation, internal workflows, AI pipelines, and n8n troubleshooting.

Why it matters: automation failures often originate in node configuration, expressions, pagination, credential assumptions, or missing error paths.

Install first if: you use Codex to build or maintain n8n workflows.

Which Codex Skill Should You Install First?

If You Are Still Looking for the Right Skill

Start with find-skills to identify relevant candidates, but treat its results as leads rather than automatic installation recommendations. Review the source and permissions of every candidate.

If You Work on GitHub Pull Requests

Start with gh-fix-ci and gh-address-comments. Add resolving-merge-conflicts if conflicting branches are a recurring bottleneck.

If You Build Web Interfaces

Use playwright-interactive for browser verification and consider design-guide when you need consistent visual rules. Vercel Agent Skills may add more targeted guidance for React and Next.js projects.

If You Ship SaaS Backends

Start with Supabase Agent Skills, security-best-practices, and sentry when those platforms match your stack. This combination covers data design, security review, and evidence-based production debugging.

If You Produce AI Media

Consider ai-video-generation for a broad video-generation workflow, hyperframes-cli for programmatic video rendering, or ai-music for recurring audio production. Check service dependencies, usage costs, licenses, and generated-asset policies before using them commercially.

If You Build Automation or Browser Workflows

Use n8n skills for structured automation and agent-browser for tasks requiring webpage interaction. Browser-control skills deserve additional review because they may interact with authenticated sessions, forms, external accounts, or sensitive page content.

If a private AI or automation workflow also needs local storage for files, datasets, or knowledge assets, a device such as the ZimaCube 2 AI NAS can provide a self-hosted storage base while Codex skills define the repeatable workflow layer.

How to Install and Audit Codex Skills Safely

Confirm the Source and Maintainer

Start with OpenAI-curated skills, official vendor repositories, or focused community projects with identifiable maintainers. Do not infer trust from the skill name alone because unrelated repositories can publish similarly named workflows.

Read SKILL.md Before Installation

Check the skill description, triggers, required tools, scripts, references, and expected outputs. A useful skill should explain when it activates and what it is allowed to do.

Inspect Scripts and Installation Commands

A SKILL.md file may be readable while an included script performs additional actions. Review shell commands, package installation, network access, file writes, credential handling, and destructive operations before execution.

Match Permissions to the Workflow

A design guide should not need production database credentials. A Reddit research workflow should not automatically need permission to publish posts. A browser skill should not receive authenticated access unless that access is necessary and explicitly intended.

Do Not Install Too Many Overlapping Skills

Skill names and descriptions help Codex determine which workflow is relevant. Installing many skills with overlapping triggers can increase selection noise and make auditing harder. Begin with three to five workflows you genuinely repeat.

Test Third-Party Skills in a Low-Risk Environment

Run a new community skill against sample data or a disposable repository before giving it production code, private files, paid services, or authenticated browser sessions. Verify its actual behavior instead of relying only on its description or leaderboard position.

Conclusion

The July 24, 2026 trend snapshot shows where short-term attention was moving: AI video generation, skill discovery, interface design, idea testing, database operations, Git conflict resolution, Reddit research, programmatic video, AI music, and browser automation.

That list is useful for discovery, but popularity should not replace workflow fit. For a stable Codex setup, begin with the smallest group of skills that solves work you already repeat. Then use trending rankings to discover specialized capabilities worth testing.

The final decision should come after reviewing the maintainer, SKILL.md, scripts, permissions, service dependencies, and recent project activity. A focused, auditable skill stack is more valuable than a large collection selected only because it was temporarily popular.

FAQ

What were the top trending Codex skills on July 24, 2026?

The supplied weekly snapshot ranked ai-video-generation, find-skills, design-guide, grill-me, prisma-client-api, resolving-merge-conflicts, reddit-automation, hyperframes-cli, ai-music, and agent-browser. The image identifies skills.sh trending data as the source.

Does the weekly ranking show the best skills for every Codex user?

No. It reflects short-term popularity or installation momentum, not universal usefulness, security, maintenance quality, or compatibility with every Codex environment.

Which skill should a new user install first?

Choose a skill that maps to a workflow you already repeat. Use find-skills if you need help discovering candidates or skill-creator if you want to turn your own established process into a reusable Codex workflow.

Which skills are useful for GitHub work?

gh-fix-ci supports failed checks, gh-address-comments supports reviewer feedback, and resolving-merge-conflicts may help when conflicting branches are a recurring problem.

Which skills are useful for AI video creation?

ai-video-generation targets a broader AI video workflow, while hyperframes-cli focuses on creating and rendering programmatic video. Their dependencies and outputs should be evaluated separately.

Are third-party Codex skills safe?

Not automatically. Review the repository, maintainer, SKILL.md, included scripts, installation commands, permissions, network access, credential requirements, and recent activity before use.

Are Codex skills the same as AGENTS.md?

No. AGENTS.md is suited to durable repository instructions, while a skill packages a reusable task workflow that may include supporting references, assets, and scripts.

Can Codex skills include scripts?

Yes. Scripts can make validation and repeated operations more reliable, but they also require closer inspection because they may read files, write data, execute commands, or access external services.