Explyt / Free instruction

Methodology for one task: six steps to a merge with no surprises in production

A printable PDF instruction from the webinar Working with AI Tools at the User Level. Pick one task from your backlog, follow six steps with checkboxes and ready prompts, fill in the result card. Written for any agent: Explyt, Claude Code, Codex, Cursor.

Get the PDF by email

One email with the PDF, then up to two follow-ups. No sales calls.

We collect your email and, optionally, your name. That is all the form sends.

Check your inbox: the PDF is on its way.

While you wait, pick a task: it has a ticket, it's testable, it touches 2–10 files, and a mistake is cheap.

What's inside

Six steps, three human checkpoints, one task

You own the checkpoints. The agent does the work in between. Each step has checkboxes and a prompt you can paste as is.

1

Project context

A one-screen AGENTS.md and two rules the agent sees before you assign the task. One chat per task.

2

Spec before code

The agent drafts task-01.md from a template: context, goal, requirements, out of scope, acceptance criteria. You approve it.

3

Tests first, locked from edits

Failing tests from the acceptance criteria, committed separately and closed to the agent before implementation starts.

4

Implement until green, review capped at 3 rounds

One prompt with hard limits: ten test runs, three review rounds, then a report of what is fixed and what is still open.

5

Your review in the right order

Test diff first, reviewer report second, code diff against the requirements third, your own test run last.

6

Capture what you learned

What the agent misread goes into AGENTS.md or a rule; a repeated manual step becomes a skill; project gotchas go into memory.

Plus: result card, symptom-based troubleshooting, and a when-to-use-what table. 8 pages.

Where it comes from

Built from the webinar and its extended guide

Sergey Pospelov from the Explyt team ran the session on 28 September 2026. The instruction condenses the slides, the live IDE demo, and the guide into steps you can follow on your own project.

Extended webinar guide

Five chapters and key terms: methodology, SDD and TDD, feedback loops, subagents and worktrees, personal setup. With the full task-01.md example and a 30-minute checklist.

Open the guide →

Webinar notes

What was said and shown, with frames from the live demo: a rule, a skill that did not load at first and why, agent onboarding, and Memory Bank in action.

Read the notes →