The skill toolbox

SDS

Structure a messy ask.

Reach for this when: you know what you want but the prompt will not come out.

Speak or type a brain dump and it sorts your own words into the three steps every good prompt has: Set the stage, Define the task, Specify the rules. Anything it guessed gets flagged so you can correct it in seconds. The steps themselves are taught in How to prompt.

The Translator

SDS on steroids.

Reach for this when: the job is too big for one prompt and needs a proper brief.

It interviews you about the work, names what good looks like, and turns your answers into a complete build prompt for an AI agent. Where SDS tidies one ask, the Translator writes the whole brief. It has its own page: how it works and when to use it.

Style-Guide Generator

Lock the look before you build.

Reach for this when: a project needs to look and sound like one thing, not five.

The tool I reach for most days. It generates a locked style guide for a project: the colours, the type, the voice, the rules everything after it must follow. Once the guide exists you stop re-describing taste in every prompt and start pointing at the file instead.

TLDR

Compress anything to its essence.

Reach for this when: something is too long and you need the point, fast.

A long reply, a dense document, a rambling thread: it hands back only what matters, no preamble, no repeating the long version. The test of a good TLDR is that you never need to go back for the original.

Red Team

Attack your own plan before reality does.

Reach for this when: you are about to commit and a miss would be expensive.

It assumes the plan has already failed and works out why. Kill-shots ranked, shaky assumptions named, and a plain verdict: proceed, proceed with fixes, or do not. The more attached you are to the idea, the more this one earns.

Moonshot

The creative optimist.

Reach for this when: every idea on your list is a 10% improvement.

It generates the ambitious 10x versions of an idea before you settle for the safe one. Use it at the start of a piece of work, while the space is still open, then hand its best output to Red Team before you commit anything real.

deep-thinking

Get the hard call right before you commit.

• Use it when: the question is hard, high stakes or open ended, and being right matters more than being fast. Strategic calls, “should I do X”, a failure that keeps coming back.
• Especially when: the obvious answer feels too easy, several answers look plausible, or you can afford a long or overnight run.
• You get: a clear verdict with a confidence level, the next steps, and what would change the answer.

consultant

Structure the mess, answer first.

• Use it when: a problem is messy or tangled and you want it structured and answered the way a top strategy consultant would: the mess broken into clean parts, the real question found, a sharp answer-first recommendation.
• Especially when: you are overwhelmed by a knot of considerations, or a decision has too many moving pieces to hold in your head.
• You get: the recommendation up front, the three reasons holding it up, and the one thing that would change the answer.

prd

Write the brief before you build.

• Use it when: you are about to build something (an app, a feature, a tool, an automation) and want a clear brief before anyone, or any AI, starts building.
• Especially when: you are handing the build to an AI coding agent, past builds over-engineered or quietly grew in scope, or “I’ll explain it as I go” keeps producing a mess.
• You get: a Product Requirements Document (PRD): a spec clear enough that a builder who has never spoken to you could build the right thing without guessing.

build

Ship it in slices, without spiralling.

• Use it when: you are about to build with an AI (Claude Code, Cursor, Cowork) and want it to end up working, not spiral into a mess.
• Especially when: past builds went off the rails, the AI over-built or quietly broke things, or it said “done” but it wasn’t.
• You get: the execution partner to prd. Plan, build a small slice, see it work with your own eyes, repeat. prd writes the brief, this ships it.

the-loop

Iterate to a bar, then stop.

• Use it when: you have a draft or output that needs to be genuinely good, not just done, and you want it refined through real cycles instead of stopping at the first attempt.
• Especially when: quality matters more than speed, first drafts are not enough, and “make this actually good” is the job. Works on writing, plans, designs, code.
• You get: an explicit quality bar, the work iterated against it until it clears, and a stop the moment it does. No slop, no polishing forever.