# Create Reliable Codex Skills

Building reusable AI skills requires clear triggers, focused guidance, and consistent files. This skill helps create, update, scaffold, and validate Codex skill packages.

## Install

```bash
npx skillstore add openai/skill-creator
```

## Metadata

- Status: approved
- Slug: skill-creator
- Skillstore revision: r2
- Version status: missing
- Tree hash: 16a6b3922c56713fb2ab0010a7a9f4e79c92123664f3d3579ed8f8875b3690c6
- Author: openai
- GitHub username: openai
- License: MIT
- Repository: https://github.com/openai/codex/tree/e38290846c11424567c9c8226f2be8a2886c5929/codex-rs/skills/src/assets/samples/skill-creator
- Ref: bc37f12c64ad89ca965c1e90d44012af0469ec4e
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: network, external\_commands, filesystem
- Quality score: 76
- Public page: https://skillstore.pages.dev/skills/skill-creator
- Manifest: https://skillstore.pages.dev/api/skills/skill-creator/manifest

## Capabilities

- Explains how to define concise skill triggers, instructions, constraints, and resource routing.
- Scaffolds new skill directories with SKILL.md and agents/openai.yaml files.
- Creates optional scripts, references, assets, and example placeholders when requested.
- Generates consistent display names, descriptions, icons, colors, and default prompts.
- Validates frontmatter fields, naming rules, description limits, and unfinished placeholders.
- Guides focused updates and independent forward testing for complex skills.

## Use Cases

- Create a First Skill: Turn a repeated workflow into a discoverable skill with focused instructions and only necessary resources.
- Standardize Team Skills: Apply consistent naming, metadata, folder structure, and validation across a shared skill collection.
- Improve an Existing Skill: Refine triggers, reduce unnecessary context, preserve policies, and add targeted resources for demonstrated needs.

## Prompt Templates

### Create a Basic Skill

```
Use $skill-creator to create a skill for [task]. Place it in [directory] and keep the instructions concise.
```

### Add Reusable Resources

```
Use $skill-creator to build a skill for [workflow]. Include [scripts, references, or assets] only where they improve repeated work.
```

### Update an Existing Skill

```
Use $skill-creator to improve the skill at [path]. Preserve user choices, invocation policy, and unrelated metadata while addressing [problem].
```

### Design and Forward Test

```
Use $skill-creator to redesign [skill path] for [complex workflow]. Validate its files and run an isolated forward test with [realistic request].
```

## Limitations

- Validation checks structure and placeholders, but it cannot prove instruction quality or task success.
- The bundled scripts create local files only and do not publish skills to a marketplace.
- Metadata generation replaces agents/openai.yaml, so existing policy or dependency fields require manual preservation.
- The skill does not install external tools, APIs, or runtime dependencies for generated skills.

## Best Practices

- Keep activation descriptions concise and specific enough to avoid unrelated requests.
- Add scripts, references, or assets only when they improve a concrete repeated workflow.
- Validate structure, test changed scripts, and use isolated forward tests when complexity warrants them.

## Anti Patterns

- Do not duplicate detailed reference content inside SKILL.md.
- Do not add placeholder directories or auxiliary documentation without a clear workflow need.
- Do not overwrite existing invocation policies, dependencies, or user choices during focused updates.

## Security Audit

- Audited at: 2026-08-17T05:39:21.669\+00:00
- Summary: All 70 static findings are false positives caused by SVG decimals, documentation examples, error text, or Markdown formatting. One medium issue remains because the metadata generator follows symbolic links when overwriting agents/openai.yaml.

## Stats

- Views: 414
- Downloads: 42
- Favorites: 3
- Popularity score: 0