Writing Task Descriptions a Coding Agent Can Follow
Describe coding tasks with goal, context, constraints and done criteria so agents build the right thing
A taste of a lesson
I asked 'add search to the products page' and the agent rewrote half the page and added a new library. What should I have written?
The agent made reasonable guesses because the request left everything open. Try: 'Add a text search box above the product list in products/ProductList.tsx that filters the already loaded products by name, case insensitive, as the user types. Do not add new dependencies or change the list layout. Add a test for matching and no results. Run the existing tests.' Notice the goal, location, constraints, scope and verification. Your turn: did you want search to query the server, or just filter what is already on the page?
Written by the teacher as an example. In your lesson the tutor answers your own questions, and like any AI it can be wrong.
What you will be able to do
- Write task descriptions with goal, context, constraints, scope and verification
- Turn vague requests into concrete ones using examples instead of adjectives
- Split large work into reviewable tasks with one concern each
- Tell an agent what not to change and what to do when something is unclear
Lesson plan
- 1 Why vague requests fail See how missing information turns into wrong guesses. Start
- 2 Goal and context Explain what should be true when done and where to start. Start
- 3 Constraints and boundaries Say what rules apply and what must not change. Start
- 4 Defining done Give the agent a way to check its own work and yours. Start
- 5 Sizing and sequencing tasks Break work into pieces that are easy to review. Start
Try asking
About this tutor
For developers, tech leads and product people handing tasks to coding assistants and agents. Most disappointing results start with a vague request. You will learn to write task descriptions the way you would brief a capable new teammate: the goal and why it matters, where in the code to look, constraints and conventions, what not to touch, how to verify the result, and what to do when something is unclear. We practise rewriting weak requests into strong ones and compare results, so you see the difference for yourself.
Reviews
4.5
2 ratingsSample
- Leila H.Sample
The before and after rewrites were the most useful thing I have learned about coding agents. My first try success rate went up noticeably.
- Bogdan S.Sample
Simple framework I now use for tickets for humans too. Wanted a few more examples for frontend work.
About the teacher
I teach developers and beginners the daily habits that make AI coding assistants useful instead of risky
9 tutors 411 lessons taught Sample
I am a software developer who teaches the working habits around AI coding assistants: how to describe a task, how to pair with an assistant without losing the thread, how to use a terminal based coding agent, and how to write project instruction files that keep it on track. I teach from the keyboard. We open a small project, try...
See Jude's profile and tutorsMore like this
Other tutors on the same or nearby topics.