CTO Toolkit

Execution Contract

The agreement between the accountable human and whoever implements — reviewed once, at the level of strategy.

How to use this

Complete before implementation begins on anything non-trivial. Approve the contract rather than approving each individual action. If you cannot fill in Intent, Risk and Tests, you are aiming at nothing — and you will hit it. If you cannot fill them in, the work is not ready to start.

The contract

  • Intent — the outcome being pursued, stated without prescribing a solution
  • Scope — what is included, and explicitly what is not
  • Constraints — architecture, security, data, performance, compatibility
  • Risk — what could go wrong and what is hard to reverse
  • Tests — how correctness will be demonstrated
  • Acceptance criteria — testable conditions for done
  • Execution plan — sequence, affected components, approval points

Autonomy declaration

  • Which autonomy level applies to this change (A0–A4)
  • Which actions require human approval before execution
  • Which systems and data the implementer may touch
  • What the implementer must never do without escalating

Quality bar

  • A reviewer can tell whether the right thing is being built, not only whether it will work
  • At least one serious alternative was considered and rejected on record
  • Irreversible decisions are flagged explicitly
  • Test strategy exists before implementation, not after

When to skip it

  • Trivial changes already covered by autonomy level A1
  • Time-boxed spikes whose output is knowledge, not production code
  • Work fully specified by an existing standard or template

This is an original, generic framework for reuse. It is not proprietary employer material.

← All toolkit resources

Related: Engineering Playbook