1. Official OpenAI Codex Skills

Start here. These are the Skills coming directly from OpenAI’s official catalog. OpenAI’s repo separates system skills, curated skills, and other installable skills, and the official docs also explain how skills, plugins, MCP, AGENTS.md, subagents, and hooks fit together.

skill-creator

https://github.com/openai/skills/tree/main/skills/.system/skill-creator

The most important Skill if you want to build your own Codex operating layer.

It teaches Codex how to create effective Skills: concise instructions, proper frontmatter, progressive disclosure, scripts, references, assets, and validation. If your team wants Codex to learn internal processes, this is where you start.

plugin-creator

https://github.com/openai/skills/tree/main/skills/.system/plugin-creator

For turning reusable Codex workflows into installable plugins.

Useful when a Skill becomes more than a personal workflow and needs to be packaged for a team, shared internally, or distributed as part of a broader Codex setup.

skill-installer

https://github.com/openai/skills/tree/main/skills/.system/skill-installer

The utility layer for installing Skills from GitHub paths and curated catalogs.

Not glamorous, but essential if you are serious about building a Codex environment instead of manually copying folders around.

openai-docs

https://github.com/openai/skills/tree/main/skills/.system/openai-docs

Gives Codex structured access to OpenAI documentation workflows.

Good for developers building with OpenAI APIs, agents, models, tools, or Codex itself.

migrate-to-codex

https://github.com/openai/skills/tree/main/skills/.curated/migrate-to-codex

Useful if you are coming from Claude Code, Cursor rules, custom scripts, or old agent instructions and want to migrate your workflow into Codex-native conventions.

2. Planning, execution, and repo workflows

This is where Codex becomes more than a code generator.

The best developers do not want an agent that writes random files. They want an agent that scopes work, plans, asks for approval, executes in batches, verifies changes, and keeps the repo clean.

create-plan

https://github.com/ComposioHQ/awesome-codex-skills/tree/master/create-plan

Simple, but extremely useful.

Before Codex starts changing files, this Skill makes it create a clear execution plan. Great for larger tasks, migrations, refactors, and anything where “just start coding” is the wrong move.

gh-address-comments

https://github.com/openai/skills/tree/main/skills/.curated/gh-address-comments

For addressing GitHub PR review comments using the gh CLI.

This is exactly the type of workflow Skills are made for: inspect comments, summarize what needs to be fixed, ask which threads to address, then apply changes.

gh-fix-ci

https://github.com/openai/skills/tree/main/skills/.curated/gh-fix-ci

One of the most practical Codex Skills.

It inspects failing GitHub Actions checks, fetches logs, summarizes what broke, proposes a fix plan, and only implements after approval. CI debugging is not creative work. It should be structured, repeatable, and fast.

pr-review-ci-fix

https://github.com/ComposioHQ/awesome-codex-skills/tree/master/pr-review-ci-fix

A more automated PR review plus CI-fix loop.

Useful for teams that want Codex to move from “here is what failed” into “review, repair, verify, and report.”

codebase-migrate

https://github.com/ComposioHQ/awesome-codex-skills/tree/master/codebase-migrate

For large codebase migrations and multi-file refactors.

The right use case: framework upgrades, API changes, folder restructuring, dependency migrations, naming migrations, or anything that touches many files and needs to be reviewable in batches.

codebase-recon

https://github.com/yujiachen-y/codebase-recon-skill

A very smart idea: understand the codebase through git history before reading code.

It surfaces hotspots, bug-prone files, momentum, ownership risk, and high-risk parts of the repo. Great before starting a refactor or joining an unfamiliar codebase.

brooks-lint

https://github.com/hyhmrright/brooks-lint

Code review grounded in classic software engineering books.

Not for everyone, but interesting for teams that want architecture, technical debt, and test quality reviewed through a more opinionated engineering lens.

Bernstein

https://github.com/sipyourdrink-ltd/bernstein

Not exactly a Codex Skill, but worth knowing.

Bernstein is a multi-agent orchestrator that can run Codex, Claude Code, Gemini CLI, and many other coding agents in isolated git worktrees with quality gates. If single-agent workflows feel too limited, this is the kind of orchestration layer developers should study.

3. Frontend, browser automation, and design implementation

Frontend work is where coding agents often fail in the most visible way.

They produce something that technically works but feels generic, misaligned, untested, or nothing like the Figma file.

These Skills help Codex work closer to the real frontend workflow.

figma-implement-design

https://github.com/openai/skills/tree/main/skills/.curated/figma-implement-design

Turns Figma designs into production-ready application code.

It is built around a structured workflow: extract Figma context, capture a screenshot, download assets, map design tokens, reuse existing components, and validate visual parity. This is one of the most important Codex Skills for frontend teams.

figma-generate-design

https://github.com/openai/skills/tree/main/skills/.curated/figma-generate-design

Useful when the goal is to generate or update design screens, not just implement them in code.

Good for product teams that want Codex to help with design exploration, wireframes, and early UI directions.

figma-create-design-system-rules

https://github.com/openai/skills/tree/main/skills/.curated/figma-create-design-system-rules

Very underrated.

This helps turn design system logic into agent-readable rules. If you want Codex to stop inventing random spacing, colors, and component patterns, give it the design system as rules.

playwright

https://github.com/openai/skills/tree/main/skills/.curated/playwright

Real browser automation from the terminal.

It lets Codex navigate, click, fill forms, capture snapshots, take screenshots, debug UI flows, and use stable element refs. Essential for any team building web apps with Codex.

webapp-testing

https://github.com/ComposioHQ/awesome-codex-skills/tree/master/webapp-testing

A practical testing Skill for real user flows.

Good for checking whether UI changes actually work in the browser instead of only trusting static code output.

screenshot

https://github.com/openai/skills/tree/main/skills/.curated/screenshot

Small but useful.

Gives Codex a cleaner way to capture and reason over visual output. Especially useful when paired with Playwright, Figma implementation, or UI QA workflows.

4. Backend, infrastructure, and deployment

The best Codex workflows are not just “write the feature.”

They are “write it, test it, deploy it, verify it, and tell me what changed.”

These are the Skills that move Codex closer to real shipping.

mcp-builder

https://github.com/ComposioHQ/awesome-codex-skills/tree/master/mcp-builder

For building and evaluating MCP servers.

If you are building agent tools, internal integrations, or custom connectors, MCP is becoming the common interface layer. This Skill gives Codex a more structured way to build those servers.

cli-creator

https://github.com/openai/skills/tree/main/skills/.curated/cli-creator

One of my favorite concepts.

It helps Codex create durable command-line tools from APIs, SDKs, curl examples, docs, scripts, or admin workflows. This is how repeated manual workflows become reusable internal tooling.

vercel-deploy

https://github.com/openai/skills/tree/main/skills/.curated/vercel-deploy

Deploy apps and websites to Vercel.

Great for preview deployments, quick testing, landing pages, prototypes, and frontend products where you want Codex to ship a working link.

cloudflare-deploy

https://github.com/openai/skills/tree/main/skills/.curated/cloudflare-deploy

For Cloudflare Workers, Pages, KV, D1, R2, Durable Objects, AI Gateway, Vectorize, Workers AI, and more.

Cloudflare’s ecosystem is powerful but wide. A Skill that helps Codex choose the right product and workflow is very useful.

netlify-deploy

https://github.com/openai/skills/tree/main/skills/.curated/netlify-deploy

Deployment workflow for Netlify projects.

Useful for static sites, frontend apps, preview environments, and teams already using Netlify as part of their delivery pipeline.

deploy-pipeline

https://github.com/ComposioHQ/awesome-codex-skills/tree/master/deploy-pipeline

More end-to-end than a single deploy Skill.

Useful for release pipelines that touch Stripe, Supabase, Vercel, verification, and rollback-style workflows.

5. Security and production safety

This is the category most developers will ignore until the first serious issue.

Do not.

If you are letting agents touch real codebases, auth flows, production infrastructure, APIs, or customer data, security Skills should be part of your default setup.

security-threat-model

https://github.com/openai/skills/tree/main/skills/.curated/security-threat-model

Repository-grounded threat modeling.

It helps Codex identify trust boundaries, assets, attacker capabilities, abuse paths, and mitigations. Good for auth-heavy apps, SaaS products, fintech, AI agents, developer tools, and anything internet-facing.

security-best-practices

https://github.com/openai/skills/tree/main/skills/.curated/security-best-practices

A general security guidance layer.

Useful when you want Codex to check implementation decisions against secure coding patterns before merging.

security-ownership-map

https://github.com/openai/skills/tree/main/skills/.curated/security-ownership-map

Useful for larger teams.

Security issues are not just technical. They are ownership problems. This Skill helps map who owns what, where security responsibility lives, and which systems need clearer accountability.

sentry

https://github.com/openai/skills/tree/main/skills/.curated/sentry

Read-only observability through Sentry.

Great for inspecting production errors, summarizing recent issues, and connecting stack traces back to the codebase.

sentry-triage

https://github.com/ComposioHQ/awesome-codex-skills/tree/master/sentry-triage

A more workflow-oriented version for diagnosing Sentry issues.

Useful when you want Codex to go from “there is an error” to “here is where it happens, here is the likely cause, here is the fix path.”

datadog-logs

https://github.com/ComposioHQ/awesome-codex-skills/tree/master/datadog-logs

For teams using Datadog.

Lets Codex filter logs from the shell, produce JSON-friendly output, and create digest workflows. Very useful when debugging production systems.

6. Product, project management, and team workflows

A lot of engineering work does not live inside code.

It lives in Notion, Linear, Jira, Slack, GitHub Issues, docs, tickets, meeting notes, specs, and messy internal context.

These Skills help Codex work where the actual work starts.

connect

https://github.com/ComposioHQ/awesome-codex-skills/tree/master/connect

The biggest unlock from Composio’s Codex Skills repo.

It connects Codex to apps through the Composio CLI so Codex can take real actions across Gmail, Slack, GitHub, Notion, Linear, Jira, Google Sheets, HubSpot, Salesforce, and many more tools.

This is where Codex stops only generating text and starts operating across the stack.

connect-apps

https://github.com/ComposioHQ/awesome-codex-skills/tree/master/connect-apps

Companion Skill for wiring up app connections and kicking off workflows from the shell.

Useful if your Codex workflows involve more than one external SaaS tool.

linear

https://github.com/openai/skills/tree/main/skills/.curated/linear

For teams using Linear.

Lets Codex interact with issues, projects, and team workflows in a more structured way.

issue-triage

https://github.com/ComposioHQ/awesome-codex-skills/tree/master/issue-triage

For Linear or Jira backlog triage.

Good for bug sweeps, prioritization, tagging, categorization, and turning messy issue lists into an actual plan.

notion-spec-to-implementation

https://github.com/openai/skills/tree/main/skills/.curated/notion-spec-to-implementation

Very practical.

Turns Notion specs into implementation plans, tasks, and progress tracking. Perfect for teams where product thinking happens in Notion and execution happens in GitHub.

notion-research-documentation

https://github.com/openai/skills/tree/main/skills/.curated/notion-research-documentation

For synthesizing multiple Notion sources into briefs, reports, or comparisons.

Useful for product discovery, technical specs, internal documentation, and research-heavy workflows.

notion-knowledge-capture

https://github.com/openai/skills/tree/main/skills/.curated/notion-knowledge-capture

Turns chats, notes, or rough ideas into structured Notion pages.

Good for saving knowledge after debugging sessions, planning calls, or product discussions.

meeting-notes-and-actions

https://github.com/ComposioHQ/awesome-codex-skills/tree/master/meeting-notes-and-actions

For turning transcripts into decisions, summaries, and owner-tagged action items.

Not strictly a developer Skill, but extremely useful for engineering managers, founders, tech leads, and product teams.

support-ticket-triage

https://github.com/ComposioHQ/awesome-codex-skills/tree/master/support-ticket-triage

For support-heavy products.

Categorizes tickets, assigns priority, suggests next actions, and drafts replies. Very useful when engineering needs to connect customer pain with product fixes.

7. Data, notebooks, research, and analysis

This is where Codex becomes useful beyond app code.

A lot of developer work today is analysis: logs, notebooks, datasets, benchmarks, dashboards, experiments, and model evaluations.

jupyter-notebook

https://github.com/openai/skills/tree/main/skills/.curated/jupyter-notebook

Creates and edits clean, reproducible Jupyter notebooks.

Great for experiments, tutorials, model evaluations, analytics, or teaching-oriented walkthroughs.

spreadsheet-formula-helper

https://github.com/ComposioHQ/awesome-codex-skills/tree/master/spreadsheet-formula-helper

For spreadsheet formulas, pivots, array formulas, and debugging sheet logic.

Useful for operators, growth teams, finance workflows, analytics teams, and technical founders who still live partly in spreadsheets.

langsmith-fetch

https://github.com/ComposioHQ/awesome-codex-skills/tree/master/langsmith-fetch

Pulls LangSmith project and test data for analysis.

Very useful if you are building LLM apps and need Codex to inspect traces, evaluations, or test outcomes.

developer-growth-analysis

https://github.com/ComposioHQ/awesome-codex-skills/tree/master/developer-growth-analysis

Analyzes Codex chat history for coding patterns and learning gaps.

Interesting for personal improvement, engineering coaching, and understanding how developers actually use agentic tools over time.

content-research-writer

https://github.com/ComposioHQ/awesome-codex-skills/tree/master/content-research-writer

Research and draft content with sourced citations.

Useful for technical blogs, changelogs, developer education, product updates, and founder-led content.

competitive-ads-extractor

https://github.com/ComposioHQ/awesome-codex-skills/tree/master/competitive-ads-extractor

More marketing than coding, but very useful for developer-led companies.

Analyzes competitor ads and extracts structured insights. Good for SaaS, devtools, AI products, and growth teams.

8. Writing, changelogs, and developer communication

Developers do not just write code.

They write PR descriptions, changelogs, docs, launch notes, internal updates, onboarding docs, READMEs, and technical explanations.

changelog-generator

https://github.com/ComposioHQ/awesome-codex-skills/tree/master/changelog-generator

Turns commits or summaries into clear changelogs.

Simple, useful, and something every product team should automate.

internal-comms

https://github.com/ComposioHQ/awesome-codex-skills/tree/master/internal-comms

For internal announcements, stakeholder updates, and engineering communication.

Useful when you need to explain what changed, what shipped, what broke, or what needs attention.

email-draft-polish

https://github.com/ComposioHQ/awesome-codex-skills/tree/master/email-draft-polish

For drafting, rewriting, and condensing emails.

Not every Skill needs to be deeply technical. Some just remove daily friction.

unslop

https://github.com/MohamedAbdallah-14/unslop

Removes AI writing patterns from text.

Useful for docs, READMEs, blog posts, changelogs, and anything where you do not want the output to sound like it was generated by a default AI assistant.

9. Meta-libraries and collections worth watching

If you want to explore beyond this list, these are the places to look.

OpenAI Skills Catalog

https://github.com/openai/skills

The official Codex Skills catalog.

Start here if you want the most reliable baseline.

Composio Awesome Codex Skills

https://github.com/ComposioHQ/awesome-codex-skills

The most practical community collection right now.

Strong focus on workflows that connect Codex to real tools, apps, CI, support, research, docs, and automation.

VoltAgent Awesome Agent Skills

https://github.com/VoltAgent/awesome-agent-skills

A broader agent-skills directory compatible with Codex, Claude Code, Gemini CLI, Cursor, Copilot, OpenCode, Windsurf, and more.

Good if you want to see where the whole skills ecosystem is heading.

AuraKit

https://github.com/smorky850612/Aurakit

An all-in-one agent skill framework with modes, subagents, lifecycle hooks, and security layers.

More complex, but worth exploring if you are building advanced agent workflows.

Vibe-Skills

https://github.com/foryourhealth111-pixel/Vibe-Skills

A broader AI skills package for staged, test-driven, context-managed work.

Worth watching, especially if you are interested in turning agent work into a governed development process.

How I would set up Codex Skills today

If I were setting up Codex for a serious developer workflow, I would not install 100 Skills.

That defeats the point.

I would start with this stack:

Core operating layer skill-creator plugin-creator create-plan skill-installer

Daily engineering workflow gh-fix-ci gh-address-comments playwright webapp-testing sentry security-threat-model

Frontend and product workflow figma-implement-design figma-create-design-system-rules notion-spec-to-implementation linear

Shipping layer vercel-deploy cloudflare-deploy netlify-deploy mcp-builder cli-creator

Team automation layer connect issue-triage meeting-notes-and-actions support-ticket-triage

That is enough to change how a developer works every day.

The point is not to collect Skills.

The point is to make Codex repeat the best version of your workflow.

Again and again.

Credits and sources

This list was curated from OpenAI’s official Codex documentation and Skills catalog, Composio’s Awesome Codex Skills repository, unicodeveloper’s Codex Skills article, Composio’s Codex Skills guide, AGENTS.md, VoltAgent’s broader agent-skills directory, and individual GitHub repositories linked above.

The Codex Skills ecosystem is moving quickly, so always check each repo before installing anything in a production environment.

Primary sources:

OpenAI Codex Skills Docs https://developers.openai.com/codex/skills

OpenAI Codex CLI Docs https://developers.openai.com/codex/cli

OpenAI Skills Catalog https://github.com/openai/skills

Composio Awesome Codex Skills https://github.com/ComposioHQ/awesome-codex-skills

Composio: Top Codex Skills https://composio.dev/content/top-codex-skills

unicodeveloper: 9 Must-Have Skills for Codex in 2026 https://medium.com/@unicodeveloper/9-must-have-skills-for-codex-in-2026-b5124b375eec

AGENTS.md https://agents.md/

VoltAgent Awesome Agent Skills https://github.com/VoltAgent/awesome-agent-skills

Bernstein https://github.com/sipyourdrink-ltd/bernstein