2btwist/dotagents/codex-imagegen
Claude Code and agent skill by 2btwist/dotagents
- community
Checking that the SKILL.md file is still in the repository.
Generate or edit real image files (PNG/JPEG) by shelling out to Codex CLI's bundled imagegen skill (gpt-image family, billed to the ChatGPT subscription, no API key). Use when the user asks Claude to generate, create, or edit images, illustrations, hero images, icons, OG/social images, textures, or photo-style assets, or when a task needs an image file that markup/SVG cannot produce. Not for charts or token-driven UI graphics (build those in code).
Install
npx vskill@latest install 2btwist/dotagents/codex-imagegenSKILL.md
The same SKILL.md was published earlier as 2btwist/dotagents/codex-imagegen.
Codex imagegen
Claude Code has no first-party image model. Codex CLI ships one: its bundled imagegen skill generates and edits images against the user's existing ChatGPT login. Verified against codex-cli 0.144.1 with codex login status reporting ChatGPT; re-check both if a run fails at the auth step.
Decision rule first
Prefer code over pixels when the surface allows it: theme-aware markup/SVG for infographics, dataviz, and UI graphics (they follow light/dark and stay accessible). Use imagegen for photographic/illustrative assets, textures, mascots, hero art, and anything destined for social/marketing where a bitmap is the deliverable.
Command shape
--skip-git-repo-checkis required outside git repos (scratchpad); harmless inside them.-s workspace-writeso Codex can write the file. Neverdanger-full-accessfor images.</dev/nullprevents a stdin hang.- Set an explicit Bash timeout of 300000 or more; generation takes 1 to 3 minutes per image.
Prompt requirements
Everything is natural language, there are no size/quality flags. Always state:
- "Use your imagegen skill" (anchors tool choice).
- The image content, style, and palette. Where the project fixes a palette or a tone, state the hex values and the mood in the prompt — the model has no access to the repo's design tokens.
- Size (e.g. 1024x1024, 1536x1024), quality (LOW for drafts and plumbing tests, HIGH only for finals), count, transparency if needed.
- The EXACT output path: "Save it as ./name.png in this directory."
- "Do nothing else."
For edits: give the input file path and list invariants ("change only the background; keep the subject unchanged").
After the run
Readthe output image and judge it before shipping; iterate with a follow-up edit run rather than regenerating from scratch when composition is right but details are off.- Masters are also archived under
~/.codex/generated_images/. - Cost awareness: each run consumes roughly 30k Codex agent tokens from the ChatGPT plan quota on top of the image generation itself. Batch related images into one run when practical.
If codex is missing or auth expired, say so and suggest ! codex login rather than falling back silently.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/codex-imagegen/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/2btwist/dotagents/codex-imagegen)