CARBON — Engineering ConfidenceUser Guide
CHAPTER 54 / 60 Pro

/ai-upgrade

Upgrade existing automated or manual tests for AI-first validation, including porting legacy suites into editable CARBON AI-intent tests and adding sparse local model-powered screenshot quality checkpoints.

Jay · Your AI test managerTesting operations

When to use it

Existing suites are valuable but brittle or difficult to maintain.

Before you start

Inventory useful deterministic assertions first. Add editable AI-intent and sparse model-assisted visual checks where judgment helps. Re-run converted tests before claiming equivalence or better coverage.

Try this in your coding agent.

/ai-upgrade Review tests/checkout and propose AI-first checks that preserve useful assertions.

Replace sample paths with your project. Slash names are shorthand; your host may use a plugin prefix or a skill picker. Invocation help →

ILLUSTRATIVE WALKTHROUGH

Existing checkout automation

An expanded teaching example, not a recorded or live execution. The report stays visible while the replay explains how to read it.

  1. 01Inventory
  2. 02Retain assertions
  3. 03Add AI checks
  4. 04Review

Coding-agent conversation

Inventory

I’m reviewing existing checkout automation. I’ll keep the scope and permissions explicit.

Retain assertions

Keep deterministic total and state assertions; add AI judgment only where it helps.

Add AI checks

Add context-aware layout and messaging checks, with reviewable evidence.

Review

A brittle screenshot-only check gains behavioral assertions and AI review. A revised test is not a passing test until executed.

CARBON

Example report

Illustration
Project → decisions → retained evidence
Project → decisions → retained evidence — TestBucks teaching screen

TestBucks course screen with illustrative analysis. The screenshot supplies context; the demo does not establish a newly observed bug.

  1. TestBucks project
  2. Command action
  3. Reviewable output

Illustrated outcome

A brittle screenshot-only check gains behavioral assertions and AI review.

Evidence to inspect

Preserve total assertion; add semantic layout review and explicit failure evidence.

Still untested / not established

A revised test is not a passing test until executed.

Ready. Read the full example or play its steps.

What a useful result contains

A bounded claim

For existing checkout automation, state the target, revision, environment and actual work performed. Keep planned, executed, blocked and untested work separate.

An inspectable trail

Preserve total assertion; add semantic layout review and explicit failure evidence. In a real run, retain the source, trace, screenshot or fixture behind the conclusion.

The qualification

A revised test is not a passing test until executed.

The next decision

Decide whether the evidence calls for a focused check, a reviewed fix or more context. Do not convert a plausible example into a verified finding.

Adapt the example to your project

  1. Replace existing checkout automation with a real feature, path or decision in your project.
  2. State the expected rule and its source. Include safe fixtures, constraints and a budget where appropriate.
  3. Ask for exact evidence and remaining gaps. Approve consequential actions separately.
/jay Review the /ai-upgrade result. Separate observed evidence, assumptions and untested scope, then recommend the next useful check.

Where to go next

/carbon-tests →/carbon-eval-design-review →

Content checked against the local command definition: skills/ai-upgrade/SKILL.md. Examples adapted from the homepage’s TestBucks walkthroughs. Availability is edition- and host-dependent.