The situation

You keep doing a familiar sequence: inspect a project, make one type of change, run checks, and report the result. Starting from a blank request each time is slow, yet a giant all-purpose instruction makes the work less clear.

The idea in one paragraph

A repeatable procedure is a short, named sequence for a recurring kind of work. It sets the trigger, boundaries, steps, checks, and stopping point. It can help you or an AI start consistently, but it should guide judgement rather than pretend that every project and situation is identical.

Procedure partIt answersExample shape
TriggerWhen to use itA routine review request
BoundaryWhat stays outNo external publishing
StepsWhat happens firstInspect before changing
CheckWhat proves progressRun the relevant test
StopWhen to askRisk or ambiguity appears

How it actually works

Start by observing a real task you repeat. Write only the decisions that stay the same, link to the project's existing guidance, and name the points where a person must decide. Use the procedure on a small case, then revise it from what actually happened. A tool-specific feature may package a procedure, but the underlying value is the clear workflow, not the packaging.

What this changes for you

  • Create a procedure only after you can describe a repeated pattern.
  • Give it a narrow trigger and a clear stop condition.
  • Review it when it causes confusion, repeated exceptions, or skipped checks.

Where it breaks

Procedures become harmful when they are applied outside their assumptions. They can preserve outdated habits, hide important judgement, or encourage an agent to act too broadly. Do not turn safety-sensitive, financial, data-changing, or publishing decisions into unattended routines.

Terms used on this page

Repeatable procedure: A written workflow for a recurring, bounded task.

Trigger: The condition that says a procedure is relevant.

Stop condition: A point where the procedure pauses for review or direction.

Return to Level 1 — Foundations.