Hi Codex team,
I’d like to propose Assetify as a candidate for an official built-in Skill, or at least as an example Skill that demonstrates a practical multimodal Codex workflow.
The motivation is simple: native image generation inside Codex is powerful, but there is still a gap between “the model generated a nice image” and “this is a usable production asset in an app, website, mini program, or game.” In real product work, generated visuals usually need several additional steps: prompt discipline, reference-image identity preservation, transparent-background validation, visual inspection, cropping, naming, packaging, and sometimes PNG/SVG export. Assetify is designed to turn that messy last mile into a repeatable Codex Skill.
The key idea is that a Codex Skill should not just be a prompt snippet. It can encode an operational workflow:
- Lock the user brief before generation, so Codex preserves exact subjects, styles, counts, labels, and output requirements.
- Use Codex native image generation to create either a single asset, a reference-image variant, or a unified asset board.
- Inspect the generated image before exporting, instead of blindly trusting scripts or fixed grid slicing.
- Default to transparent-background assets and verify the actual PNG alpha channel, because image models often draw fake checkerboard “transparency.”
- Slice asset boards into reusable files only after visually deciding whether equal-grid, detected boxes, or manual crop boxes are appropriate.
- Export PNG assets plus SVG wrappers when UI pipelines want SVG files, while clearly documenting that these are not pure editable vectors.
This matters because Codex is becoming more than a code assistant. It is increasingly an agentic workbench where visual generation, file operations, scripts, and implementation can live in one loop. Asset generation is a good example of that shift: the valuable output is not just an image or just code, but a production-ready artifact that can immediately enter a UI pipeline.
I see Assetify as a small but concrete example of “image generation + programming model + filesystem/tool execution” becoming one workflow. It is especially useful for:
- App and website illustration assets
- WeChat mini program feature icons
- Reference-based character variants
- Transparent PNG stickers or mascots
- Batch asset boards that need consistent style
- UI pipelines that need named PNG/SVG deliverables
I’m not opening this as a direct code submission yet. I’d first like to ask whether this direction fits the Codex roadmap for official or example Skills. If it does, I’d be happy to adapt the repository structure, naming, documentation, or scope to match the standards maintainers expect.
Repository: https://github.com/NimaChu/assetify
Hi Codex team,
I’d like to propose Assetify as a candidate for an official built-in Skill, or at least as an example Skill that demonstrates a practical multimodal Codex workflow.
The motivation is simple: native image generation inside Codex is powerful, but there is still a gap between “the model generated a nice image” and “this is a usable production asset in an app, website, mini program, or game.” In real product work, generated visuals usually need several additional steps: prompt discipline, reference-image identity preservation, transparent-background validation, visual inspection, cropping, naming, packaging, and sometimes PNG/SVG export. Assetify is designed to turn that messy last mile into a repeatable Codex Skill.
The key idea is that a Codex Skill should not just be a prompt snippet. It can encode an operational workflow:
This matters because Codex is becoming more than a code assistant. It is increasingly an agentic workbench where visual generation, file operations, scripts, and implementation can live in one loop. Asset generation is a good example of that shift: the valuable output is not just an image or just code, but a production-ready artifact that can immediately enter a UI pipeline.
I see Assetify as a small but concrete example of “image generation + programming model + filesystem/tool execution” becoming one workflow. It is especially useful for:
I’m not opening this as a direct code submission yet. I’d first like to ask whether this direction fits the Codex roadmap for official or example Skills. If it does, I’d be happy to adapt the repository structure, naming, documentation, or scope to match the standards maintainers expect.
Repository: https://github.com/NimaChu/assetify