Skip to content

Controlled change

What it is

Controlled change is CodeClone's pre-edit and post-edit verification workflow for repository modifications. It provides intent declaration, scope boundaries, structural risk assessment, and deterministic verification before and after code changes, producing a durable audit trail of what was known, what was declared, and what was verified.

Why it exists

Unreviewed, unscoped edits are the normal failure mode for both human and AI-assisted development: a change silently grows beyond its intended files, nobody notices a new dependency cycle until it ships, and "I tested it" is not something a later reviewer can verify. Controlled change exists to make scope and verification explicit and checkable rather than a matter of trust — particularly important for AI agents, which have no persistent memory of their own actions across sessions unless the workflow records it for them.

How it fits together

Controlled change is the umbrella workflow that several other concepts plug into:

Concept Role in the workflow
Blast radius Computed at intent declaration; shows the structural risk of the declared scope
Edit permission and edit_allowed The single authorization signal the workflow produces before editing may begin
Engineering Memory Consulted mid-edit for prior decisions, incidents, and stale-context warnings
Review receipts The durable, auditable record produced when the workflow finishes 
graph LR
    A["Analysis baseline"] --> B["Controlled change"]
    B --> C["Blast radius"]
    B --> D["edit_allowed"]
    B --> E["Engineering Memory"]
    B --> F["Review receipt"]

The workflow is lightweight for documentation-only changes and full-featured for structural Python or governance-config changes: a session holds exactly one trackable active intent at a time, and starting a new one before finishing the current one evicts it from tracking with no recovery except redeclaring the same scope.