AI Perks curates and provides access to exclusive discounts, credits, and deals on AI tools, cloud services, and APIs to help startups and developers save money.
What Are OpenAI Codex Skills?
Codex Skills are reusable workflow bundles that extend OpenAI Codex with task-specific capabilities. A skill packages instructions, scripts, and references so Codex executes recurring tasks the same way every time. Launched in December 2025 as an experimental feature, skills work across the Codex CLI, IDE extensions, and web app.
If you're building agent workflows with Codex, skills eliminate prompt drift and make Codex behave predictably. And if you want to power Codex without paying OpenAI directly, AI Perks provides free OpenAI API credits worth $500-$50,000+ through programs most developers don't know exist.
Save your budget on AI Credits
| Software | Approx Credits | Approval Index | Actions | |
|---|---|---|---|---|
Promote your SaaS
Reach 90,000+ founders globally looking for tools like yours
What Codex Skills Actually Do
Skills solve three pain points with AI coding agents:
| Problem | How Skills Solve It |
|---|---|
| Inconsistent agent behavior | Skills enforce step-by-step workflows |
| Repeated prompt engineering | Write the prompt once, invoke forever |
| Knowledge silos across teams | Skills are version-controlled and shareable |
Skills are essentially "agent macros" - you define a workflow once, then invoke it whenever you need it. Codex follows the skill's instructions exactly instead of improvising.
AI Perks curates and provides access to exclusive discounts, credits, and deals on AI tools, cloud services, and APIs to help startups and developers save money.
How to Create a Codex Skill
A skill is a directory containing a SKILL.md file plus optional scripts and reference documents.
Minimum Skill Structure
my-skill/
├── SKILL.md # Required: instructions and metadata
├── scripts/ # Optional: helper scripts
│ └── helper.sh
└── references/ # Optional: documentation, examples
└── docs.md
SKILL.md Format
---
name: deploy-to-staging
description: Deploys the current branch to staging environment with health checks
---
## When to Use This Skill
Use when the user asks to deploy code to staging, push to staging, or test a branch on staging.
## Steps
1. Run `git status` to verify clean working tree
2. Build the project with `npm run build`
3. Push to the staging branch with `git push origin HEAD:staging`
4. Trigger the staging deploy with `./scripts/deploy.sh`
5. Wait for health check to return 200
6. Report deploy URL to the user
## Inputs
- Current branch name (auto-detected)
- Optional: feature flag overrides
## Outputs
- Deploy URL
- Health check status
- Any deploy errors
Required Frontmatter Fields
| Field | Purpose |
|---|---|
name | Unique skill identifier |
description | What the skill does + when to use it |
The description is critical - it's what Codex uses to decide whether to invoke the skill automatically.
How to Invoke Codex Skills
There are two invocation methods:
Explicit Invocation
Type the skill name with the $ prefix:
$.deploy-to-staging
Or browse all available skills:
/skills
This forces Codex to use the named skill regardless of the prompt.
Implicit Invocation
When you give Codex a natural-language prompt that matches a skill's description, Codex picks it automatically:
You: "Deploy this branch to staging"
Codex: [Detects deploy-to-staging skill matches → invokes it]
This works because Codex's progressive disclosure loads skill names and descriptions first, then loads the full SKILL.md only when it picks a relevant skill.
Codex Skills Best Practices
1. Keep Each Skill Scoped to One Job
A skill that does too many things becomes unpredictable. Split complex workflows into multiple smaller skills.
Bad:
name: full-release-pipeline
description: Builds, tests, deploys, monitors, and notifies for releases
Good:
name: build-and-test
description: Builds the project and runs the test suite
name: deploy-to-staging
description: Deploys to staging after build/test passes
name: notify-team
description: Sends deploy notifications to Slack
2. Start With 2-3 Concrete Use Cases
Write skills based on actual workflows you do repeatedly. Don't create skills for hypothetical scenarios.
Common starter skills:
deploy-to-stagingrun-database-migrationgenerate-pr-descriptionupdate-changelogcreate-feature-branch
3. Define Clear Inputs and Outputs
Specify exactly what the skill needs and what it produces:
## Inputs
- target-environment: "staging" or "production"
- skip-tests: optional boolean (default: false)
## Outputs
- deploy-url
- deploy-duration-seconds
- error-message (if failed)
4. Write Descriptions That Match User Language
The description determines when Codex auto-invokes the skill. Use the words developers actually say.
Bad:
description: Initiates CI/CD orchestration with branch promotion
Good:
description: Deploys current branch to staging - use when user says "deploy", "push to staging", or "test on staging"
Why Codex Skills Cost Money (And How to Make Them Free)
Every Codex skill invocation consumes OpenAI API tokens like any other Codex prompt. Skills don't reduce per-invocation cost - they just make the workflow consistent.
Typical Codex Skill Cost
| Skill Complexity | Tokens Used | Cost (GPT-4.1) | Cost (GPT-4.1 Nano) |
|---|---|---|---|
| Simple (deploy script) | ~3,000 | $0.024 | $0.0012 |
| Medium (refactor + test) | ~15,000 | $0.12 | $0.006 |
| Complex (full feature impl) | ~50,000 | $0.40 | $0.02 |
A team using 20 skill invocations per developer per day spends $50-$200 per developer per month on Codex skill execution alone.
AI Perks provides free OpenAI credits that cover this cost entirely.
Get Free OpenAI Credits to Power Codex Skills
| Credit Program | Available Credits | How to Get |
|---|---|---|
| OpenAI (GPT models) | $500 - $50,000 | AI Perks Guide |
| Microsoft Founders Hub (OpenAI access) | $500 - $1,000 | AI Perks Guide |
| Azure OpenAI Service Credits | $1,000 - $50,000 | AI Perks Guide |
| AWS Activate (alternative models) | $1,000 - $100,000 | AI Perks Guide |
| Accelerator + VC Programs | $1,000 - $5,000 | AI Perks Guide |
Total potential: $4,000 - $206,000+ in free credits
At $50/developer/month in skill costs, even a $5,000 credit grant funds 8+ years of Codex skills usage for one developer or 1 year for an 8-person team.
Codex Skills vs Custom Functions vs Tools
| Feature | Skills | Custom Functions | Tools |
|---|---|---|---|
| Reusable | Yes | Yes | Yes |
| Version-controlled | Yes (git) | Yes | Depends |
| Cross-team shareable | Yes | Limited | Yes |
| Invoked by name | Yes | Yes | Yes |
| Auto-invoked by intent | Yes | No | Limited |
| Includes scripts | Yes | No | Yes |
| Best for | Workflows | Single tasks | Integrations |
Skills are the most flexible option for codifying recurring developer workflows. Functions are simpler but less powerful. Tools are for external integrations.
Step-by-Step: Build Your First Codex Skill
Step 1: Get Free OpenAI Credits
Subscribe to AI Perks to access free OpenAI credit programs. This funds your Codex skills usage at zero cost.
Step 2: Identify a Repetitive Workflow
Pick something you do at least weekly. Common candidates:
- Deploy to staging
- Run database migration
- Generate PR description from commits
- Update changelog from commits
- Run security scan
Step 3: Create the Skill Directory
mkdir -p ~/.codex/skills/my-skill
cd ~/.codex/skills/my-skill
Step 4: Write SKILL.md
Use the template from earlier. Be specific about steps, inputs, and outputs.
Step 5: Test the Skill
In the Codex CLI, run:
$.my-skill
Iterate on the SKILL.md until Codex executes the workflow consistently.
Step 6: Share With Your Team
Commit your ~/.codex/skills/ folder (or a subset) to git. Team members can pull and gain instant access to your skills.
Frequently Asked Questions
What's the difference between a Codex Skill and a regular prompt?
A skill is a versioned, reusable workflow bundle; a prompt is a one-off instruction. Skills enforce consistency across runs and team members. Use skills for any task you do repeatedly. Free OpenAI credits via AI Perks cover the API cost of skill execution.
Can Codex Skills be used in the IDE and CLI?
Yes. Skills are version-controlled, shareable across teams, and available across all Codex surfaces - Codex CLI, IDE extensions (VS Code, JetBrains), and the Codex web app. The same SKILL.md works everywhere.
Do Codex Skills cost extra?
No, skills are free to create and store. They consume normal OpenAI API tokens when invoked, just like any other Codex prompt. To eliminate that cost, use free OpenAI credits worth $500-$50,000+ from AI Perks.
How do I share Codex Skills with my team?
Commit your skills directory to git. Most teams maintain a shared repository of skills (e.g., team-skills/) that all developers clone and link to their ~/.codex/skills/ folder. Skills become part of your engineering tooling.
Can Codex Skills include shell scripts?
Yes. A skill directory can contain shell scripts, Python helpers, reference docs, or any other files. The SKILL.md instructions can reference these files and invoke them as part of the workflow.
Are Codex Skills available in 2026?
Yes, Codex Skills launched as an experimental feature in December 2025 and remain available in 2026. The feature is actively developed - check the official Codex changelog for updates.
What's the best way to learn Codex Skills?
Start by creating one skill for your most-repeated workflow (e.g., "deploy to staging"). Iterate based on Codex's behavior. Then expand to 5-10 core team workflows. Free OpenAI credits via AI Perks let you experiment without worrying about token costs.
Build Codex Skills With Zero API Costs
Codex Skills make AI coding agents predictable and shareable - but every skill invocation costs OpenAI tokens. AI Perks eliminates that cost:
- $500-$50,000+ in free OpenAI credits
- Stacking strategies for $100,000+ in combined credits
- 200+ additional startup perks beyond AI credits
- Updated programs every month
Codex Skills are powerful. Make them free with credits at getaiperks.com.
