X01Runnable examples

Three source-defined workflows. One single-file builder.

Each example is a small .NET project that links the same SuperIntelligenceBuilder.cs, declares a task as a fluent C# statement, and keeps the generated files in the repository for immediate inspection.

You can read every task and output without running a coding agent. A live rerun is optional and may replace the checked-in outputs named by that task.

X02Three workflows

Workflow	What the task produces
Specification workflow	Agent instructions, project principles, a product specification, technical plan, and dependency-ordered tasks. The documents link requirements and acceptance criteria to planned work and verification; they do not implement the CSV import feature.
Static web design workflow Repair-cafe demonstration	A repair-cafe brief, three design directions, a recorded selection, visual rules, a review checklist, and a local HTML/CSS/JavaScript demonstration.
Enterprise C# practices	Agent instructions, EditorConfig, shared MSBuild defaults, architecture and coding guidance, risk-based quality gates, and a pull-request checklist.

X03Run an example

Use the complete repository checkout. The example project files link the shared builder by a relative path, so downloading only Program.cs and a project file is not a complete runnable bundle.

1. Inspect safely

cd examples/specification-workflow
dotnet build
dotnet run

With no execution flag, the builder prints a dry-run plan and the exact fluent statement. It does not probe or launch Codex and does not write files. This preview does not establish authentication or CLI compatibility.

2. Permit execution

dotnet run -- --yes,please,proceed

All three example programs explicitly select Codex. A live run therefore needs an installed, authenticated Codex CLI as well as the .NET 10 SDK.

3. Confirm the checked-in instruction file

terminal reply

Continue: yes, please proceed
Stop:     no

Every example already contains AGENTS.md, so the first live run from this checkout also asks for confirmation. Approval permits one run; it does not approve every edit or provide rollback. Review the diff afterward. The complete behavior is documented under Execution and completion.

X04Recorded evidence

The checked-in outputs came from three authenticated builder-driven Codex runs. The example programs were built, and the generated documents, configuration, and website were checked according to their format. Those facts do not make an agent completion report independent proof that every output is correct.

See the validation record for the exact commands, traceability counts, browser checks, and remaining platform limits.