Update on what the visual docs for @ToDesktop's system flows now looks like. All this is driven from a JSON schema that LLMs read when working on related features/bugs. Prompt is below in comments if you want to make it for your app.

Skills clutter up context in Codex and most of them are irrelevant to your daily work (like the `$hatch-pet` skill) but still take up space and distract the model. The title and description of every skill is in your context. You should disable "Implicit Invocation" for most of them, so they will be removed from context unless explicitly invoked. Run this prompt in Codex (you can prob do something similar in Claude): ``` List standalone Codex skills in ~/.codex/skills and ~/.agents/skills. Exclude plugin/namespaced skills (`:`). Show title, GPT-2 BPE tokens for title+description, and implicit invocation (default yes unless openai.yaml says false). Local only; flat A–Z table. ``` Then tell Codex which ones you want to make explicit (or ask Codex to recommend for you). You can still use the skill when it's explicit but you will need to explicitly reference the skill in your prompt. I ended up making most explicit.

Codex's new Intelligence slider has got some of that Claude Code style shiny attention slop.

Fable + Claude Code is very good at working through component design/UX and highlighting pros/cons of different design options. Ask it to output an HTML artifact to explain its thoughts. Prompt + Artifact are in thread below.