Commands and direct tools

The Claude column is typed in the Claude Code conversation. The Codex column is typed in the Codex CLI conversation; $workflow-name denotes a skill, not a Unix command. Supply arguments or follow-up context in that conversation, then allow the provider to dispatch permitted tools. Nothing in the table tells a normal user to run python3 manually.

Workflow commands

Claude	Codex equivalent	Purpose
/guide	$workflow-guide	Open guidance and command help.
/demo	$workflow-demo	Set up, inspect or reset demonstration workspaces.
/analyse	$workflow-analyse	Understand a repository, document set or architecture input before judging it.
/criteria	$workflow-criteria	Establish review criteria before inspection.
/review	$workflow-review	Conduct an independent review and record findings.
/rework	$workflow-rework	Resolve recorded findings under the right separation of duties.
/diagnose	$workflow-diagnose	Root-cause analysis before remediation.
/develop	$workflow-develop	Controlled implementation or document change work.
/govern	$workflow-govern	Repository or content governance analysis.
/research	$workflow-research	Research to a decision with disciplined evidence handling.
/sabsa	$workflow-sabsa	SABSA library lifecycle.
/publish	$workflow-publish	Render reviewed semantics into enterprise deliverables.
/repository	$workflow-repository	Scoped EA-domain projection repository operations and Open Exchange interoperability.
/commercial	$workflow-commercial	Opt-in supplier opportunity-to-exit coordination with commercial-delivery/1; strict provider controls require current runtime evidence.
/layered	$workflow-layered	Consequential JAI L1–L6 flow with typed contracts, bottom-up review, bounded rework/reformulation and L0 escalation.
/fabric-report	$workflow-fabric-report	Independent audit/report view over a Fabric record.

Direct tools

The provider workflows are instruction/adaptation surfaces that may invoke direct tools under the common runtime. These tools are implementation interfaces for the agent, CI, or developer troubleshooting—not a second end-user procedure. Their availability alone does not prove that a particular run executed them: inspect tool results and evidence. See the developer reference, interaction patterns and human authority.

• tools/commercial_router.py
• tools/model_router.py
• tools/provider_parity.py
• tools/criteria.py
• tools/findings.py
• tools/publication.py
• tools/repository.py
• tools/sabsa.py
• tools/execution_record.py
• tools/native_enforcement_26104.py
• fabric/fabric.py
• demo/demo.py