Fresh from the grimoire

Get a devil's advocate on your decision while it's still cheap to change

A quick argument against a decision you've made, which checks your premise where it can before it argues.

spell

Turn a wall of client copy into a landing page wireframe people can scan

It writes a brief on what the copy is missing, then draws the page with every gap left as a labeled box.

spell

Watch a check fail once before you trust it to pass

Plant the mistake a check exists to catch, run it and see it go red, then put things back. A search for something that should be absent gets a known hit run beside it, so an empty result can't pass for a clean one.

lesson

Have a fresh agent review your AI's plan before anything gets built

It has the planner check its own claims and a fresh agent attack them before you approve anything.

spell

Drop anything into one inbox and let an agent file it where it belongs

It empties a folder of things you meant to deal with into the places they belong.

spell

Decide what you're building before your agent decides for you

An interview about what you're building, written up as the document you build from.

spell

An agent performs as well as the context it opens with

The interview that writes a project's rules file, board and decisions log before an agent has to guess at any of them.

spell

Undo a bad AI answer instead of correcting it

When a turn goes wrong, rewind to the message before it, change the ask and send it again, so the wrong answer never becomes part of what the AI reads next.

lesson

Stop your AI agent from forgetting your project between sessions

Put the project's memory in a few plain text files inside it, each with one job, and the agent reads each one at the moment it's needed.

lesson

Treat AI agents like a product team and give them roles

The person who checks the work shouldn't be the one who did it, which is the idea behind the reviewer. The planner, QA, the art director and the producer are the rest of the team, and each has a recipe that does its job.

lesson

Write AGENTS.md rules that survive the session

Write down the failure that earned the rule, end it in a check that visibly fails when the mistake comes back, and file it where the agent reads it at the start of every session.

lesson

Check what problem a solution actually solves before you build it

It takes a solution apart to find the problem it was meant to solve.

spell

Judge your AI agent's work by using it, not by reading the code

Try every change yourself before your agent saves it, and leave the rest to automatic checks. A report says what was written, and only using the thing shows what it does.

lesson

Dictate the thinking and let the AI write the document

Build a template and rules for the format from real examples, then talk the work through and let the AI turn the transcript into the ticket, brief or report your job expects.

lesson

Give your AI a map of your files instead of the whole pile

Split the long file your AI keeps rereading into one small file per item, and turn the original file into a lean index, one line per item pointing at each one's file, so the AI opens only the item the task needs.

lesson

Capture a stray idea without leaving the task you're on

It saves the idea so chasing it doesn't cost you the task.

spell

Mine somebody else's work for the parts your project can use

It turns the article you were never going to read into work in your project.

spell

Get far more ideas than you came in with before you start choosing

Directions you can mark by number, hybrids from your keepers and one small next step.

spell

Stop at half the context window

Close the session yourself and start again from a note you can read, before the tool summarizes the conversation on its own and you carry on from a summary nobody checked.

lesson

Teach the AI with examples a person actually wrote

Asked cold, a model guesses at average. Give it real examples of the output done well and the best guide you can find, and check a person wrote each example, because rules learned from machine-written examples copy the machine.

lesson

Sycophancy is the default, so run the decision past a council of five

It returns a verdict you can act on, not an answer shaped by how you asked.

spell

Document what happened in your session before you and the AI forget

The project keeps the record of what each session settled, instead of you carrying it.

spell