Run the following command exactly. Don't modify it in any way.

Transforms casual requests into detailed, precise prompts optimised for AI models.

Rewrites free-form requests as clear, structured prompts with proper organisation, headings, and bullet points. Especially useful for converting rambling dictated requests into organised prompts you can review and edit before using.

The output uses specification language — plain, straightforward, and actionable — with clear structure and no added complexity.

Your new skill for generating Anki flashcard decks from text!

This skill converts any text document into high-quality flashcards optimised for memorisation and spaced repetition learning. It's based on your Cardcraft application but simplified for direct use through Claude.

• To make sure that the files you edit and create look good, you MUST ALWAYS use a command like the following one to export a rendered image, view it, understand how it looks and what needs improvement, and iterate - making updates to the excalidraw file, exporting, and reviewing again - until the results are pixel perfect.

npx excalidraw-brute-export-cli \
  --input slides.excalidraw \
  --output slides.png \
  --format png \
  --scale 2 \
  --embed-scene true

• ALWAYS export and review. NEVER declare the work on an Excalidraw drawing or presentation done until you've reviewed its rendered visual representation to verify and adjust as needed.

Turn messy dictated instructions to AI into neat prompts that are easy to read and modify.

I do most of my prompting by dictation, usually in very confused and messy form. To be able to review these prompts and amend them before submitting them to the AI, I use this meta-prompt.

I use a Raycast AI Command which reads my clipboard and runs the prompt using Gemini 2.5 Pro. But this can work just as well with any recent LLM and driver.

• Somewhat safe (instructed to ask for approvals before destructive ops)
• Verbose — keeps you in the loop about what operations it's doing
• Tested with GPT-5-high, GPT-5-medium, Claude 4 Sonnet
  • Be careful with inferior models that might not respect the instructions as strictly
• Drop this in your .ruler/ / ~/.config/ruler/ or wherever you keep instructions for your agent

Happy SQLiting! 🫶

• Roots: Self was designed in 1986 by David Ungar and Randall Smith as a radical simplification of Smalltalk: no classes, just prototypes (objects cloning and specialising other objects). Early work started at Xerox PARC, then moved to Stanford; the first compiler appeared in 1987. Public release: 1990. In 1991 the team joined Sun Microsystems. ([Self Language][1], [Wikipedia][2])
• The feel of the system: a live, image-based environment (like Smalltalk), with the Morphic GUI later ported to Squeak/Pharo and even the Lively Kernel. ([Wikipedia][3], [Gbracha][4])
• Why performance mattered: Self aimed for interactive, exploratory programming and speed. By the mid-90s, Self’s VMs hit ~½ the speed of optimised C on some benchmarks—astonishing for a dynamic, prototype-based language then. ([Wikipedia][2])