THE PRACTICAL FIELD GUIDE
Your project.
Your questions.
Evidence you can inspect.
A visual guide to CARBON’s 60 commands. Start with one useful request, follow the walkthrough, then take the same thinking into your own coding agent.
MEET YOUR TEST MANAGER
Let Jay decide where to look first.
Jay examines the project and current changes, prioritizes consequential risks, coordinates the available sub-agents and reports back with evidence, trends and a qualified opinion on shipworthiness.
/carbonNo project handy? Create a disposable demo →
One request. In your coding agent.
Install the appropriate CARBON edition in a supported host, open the project and use its command or skill picker. This booklet does not install CARBON or run tests.
Claude
Commands commonly use the plugin prefix, such as /carbon:carbon or /carbon:jay. Use the name shown by your installed plugin.
Codex
Choose the CARBON skill, or use its supported $carbon form. Names and invocation UI are host-controlled.
Other configured hosts
Use the installed integration’s prompt, command or tool picker. A slash shorthand in this guide is not a universal terminal command.
Your access, your limits
Keep your existing frameworks. Your host supplies tools and model access; model usage, credentials and permissions remain under that host’s policies.
Read results, not just green checks.
Observed
A check ran, with a recorded outcome and evidence.
Proposed
A test or fix is suggested. It has not yet established anything.
Blocked / untested
Access, tools or scope prevented execution. Missing evidence is not a pass.
Qualified confidence
A judgment tied to evidence, assumptions and known gaps—not a guarantee of defect-free software.
CHOOSE YOUR NEXT QUESTION
The complete command directory.
Core includes 12 shared command chapters; Pro includes all 60. /j is an alias for /jay, not an extra workflow. Individual chapter notes explain edition-specific limits.
Start & investigate
Core + Pro/carbon
You have a change to validate but do not yet know where testing effort belongs.
Read + watch →Core + Pro/jay
You want a testing conversation: what matters, what happened, or what to investigate next.
Read + watch →Core + Pro/full
You need a coordinated campaign rather than a bounded first look.
Read + watch →
Core + Pro/carbon-test
You need a strict boundary around one feature or behavior.
Read + watch →
Pro/carbon-functionality
You know the behavior to verify, including its edge and failure cases.
Read + watch →
Pro/carbon-browser
The answer depends on interacting with a rendered page.
Read + watch →Core + Pro/carbon-issues
You want bug discovery, including surprising multi-step journeys.
Read + watch →Core + Pro/carbon-demo
You have no project handy, or want to learn the evidence loop safely.
Read + watch →Pro/carbon-more
A previous assessment left useful checks unexecuted.
Read + watch →Core + Pro/carbon-help
You are unsure which workflow matches the problem.
Read + watch →Experience & accessibility
Core + Pro/carbon-accessibility
You need contextual accessibility evidence beyond simple rules.
Read + watch →
Pro/carbon-ui
You suspect clipping, overlap, visual inconsistency or broken responsive states.
Read + watch →Pro/carbon-ux
A sequence works technically but feels confusing or untrustworthy.
Read + watch →Pro/carbon-usability
You want to know whether a person can find and complete a task.
Read + watch →Pro/carbon-personas
Different user goals or contexts may reveal overlooked friction.
Read + watch →
Pro/carbon-compatibility
A feature must work across specific environments.
Read + watch →
Pro/carbon-intl
You are preparing the architecture for more markets.
Read + watch →Pro/carbon-localization
Translations exist but may be wrong in product context.
Read + watch →
Pro/carbon-content
Copy, links or product claims might mislead users.
Read + watch →
Pro/carbon-geo
You want to improve the evidence a search or answer system can find.
Read + watch →Security & resilience
Pro/carbon-security
You need a scoped review of application trust boundaries.
Read + watch →
Pro/carbon-security-review
AI inputs, retrieval or tool calls introduce new trust boundaries.
Read + watch →
Pro/carbon-privacy
You need to understand where personal data goes.
Read + watch →
Pro/carbon-state
The problem appears after refresh, navigation, sign-in or concurrent activity.
Read + watch →
Pro/carbon-reliability
You care about recovery and consistent behavior when dependencies fail.
Read + watch →
Pro/carbon-performance
An interaction is slow and you need attributable measurements.
Read + watch →Pro/carbon-load
You need load readiness or a controlled capacity investigation.
Read + watch →Pro/carbon-stress
You need to understand behavior beyond normal operating conditions.
Read + watch →Pro/carbon-networking
Offline, retry, caching or timeout behavior is uncertain.
Read + watch →Pro/carbon-api
A contract boundary needs focused input/output validation.
Read + watch →Pro/carbon-data
A value changes incorrectly across computation or persistence.
Read + watch →Pro/failure
You need to force errors without damaging the original project.
Read + watch →Proof & release confidence
Pro/proof
A consequential code property has a small, supported formal scope.
Read + watch →Pro/carbon-confidence
You have evidence but still need a reasoned shipping judgment.
Read + watch →Pro/carbon-confidence-init
A project lacks a durable confidence and evidence structure.
Read + watch →Pro/carbon-confidence-plan
An AI change needs a risk-led validation design before execution.
Read + watch →
Pro/carbon-review
You want a fresh challenge to AI-generated work or supporting evidence.
Read + watch →Pro/carbon-skeptical-review
A readiness claim feels stronger than the evidence.
Read + watch →Pro/carbon-release
You need a scoped recommendation to ship, canary, hold or roll back.
Read + watch →Pro/carbon-release-review
A release recommendation needs an independent challenge.
Read + watch →Pro/carbon-incident
An AI feature has produced a real harmful or incorrect outcome.
Read + watch →AI evaluations
Pro/carbon-eval
You need evidence about model, prompt, retrieval or agent behavior.
Read + watch →Pro/carbon-eval-design-review
The evaluation itself may be measuring the wrong thing.
Read + watch →Pro/carbon-statistical-review
You want to know whether an apparent evaluation improvement is credible.
Read + watch →Pro/carbon-agentic
Other AI agents need to discover and operate your interface safely.
Read + watch →Pro/chatbot
You need a broad conversational test matrix or multi-turn suite.
Read + watch →Pro/bias
Uneven representation or outcomes could be hidden by the aggregate.
Read + watch →Testing operations
Pro/carbon-tests
You want reusable test intent rather than one-off chat findings.
Read + watch →Pro/carbon-generate
You need prioritized test candidates from a change or verified defect.
Read + watch →Core + Pro/carbon-map
You need to see journeys, risk and gaps spatially.
Read + watch →Pro/carbon-full-report
You need the complete retained quality record in one place.
Read + watch →Pro/carbon-stats
You need an evidence-qualified view of testing activity and trends.
Read + watch →Pro/carbon-integrations
Your issue tracker or test manager holds relevant quality evidence.
Read + watch →Pro/ai-upgrade
Existing suites are valuable but brittle or difficult to maintain.
Read + watch →Pro/carbon-auto
A bounded repeated check loop fits a changing local feature.
Read + watch →Core + Pro/carbon-background
You want opt-in checks while working between changes.
Read + watch →Pro/carbon-forever
You explicitly want ongoing discovery across successive coverage gaps.
Read + watch →Core + Pro/carbon-settings
You need to inspect or change local execution defaults.
Read + watch →Pro/carbon-human
Business intent or risk tolerance cannot be inferred from code.
Read + watch →Core + Pro/carbon-pro
You want to compare deeper capabilities and support.
Read + watch →Use it as a booklet.
Every chapter is a real HTML page with working relative links. Keep the folder together for offline reading. Search the command sidebar, move through chapters, or open the complete printable edition.
The TestBucks screens come from teaching fixtures. Animated chat and report outcomes are authored illustrations, not fresh runs. No subscribers, accounts or external services are accessed.