SECTOR 02 // HOW-TO GUIDES

Author scenarios with Claude Code

Use the bundled writing-resilience-scenario skill so Claude Code drafts and validates scenarios against the real engine contract instead of guessing.

Goal: write a correct scenario faster by letting Claude Code lean on a skill that already knows the lifecycle, the step shape, and the validation rules.

What ships with the repository #

The repository includes a Claude Code skill at skills/writing-resilience-scenario/. It is a distilled operating manual for authoring: the scenario lifecycle, the step envelope, the result namespaces, the findings ledger, and an annotated template that validates clean. It exists so Claude Code writes scenarios against the real contract rather than reverse engineering it from the Go source.

Use it in this repository #

Open the repository in Claude Code and the skill loads on its own. Ask for a scenario in plain language:

Add a scenario that kills the database mid-write and asserts the API recovers.

Claude loads the skill when the request matches, or you can invoke it directly:

/writing-resilience-scenario

Either way it drafts a kind: Scenario file, wires the providers and builtins, and runs the validate loop until the file is well formed.

Use it on your own project #

The skill is a directory, so it travels. To make it available while working on a project outside this repository, symlink it into that project’s .claude/skills/, or into your personal ~/.claude/skills/ to have it everywhere:

ln -s /path/to/shinari/skills/writing-resilience-scenario \
  ~/.claude/skills/writing-resilience-scenario

The authoritative detail lives here #

The skill carries a condensed working set. For the full detail behind anything it drafts, read the reference:

Treat a clean shinari validate as the definition of well formed, whether you wrote the scenario by hand or Claude Code drafted it.