| name | scenario-writer |
| description | Picks a new scenario to create based on the description of the user, delegates the writing of it to the ai-scenario-writer subagent and verify the correctness. |
Implement the scenario proposed by the user if that has not been implemented yet. Then invoke the ai-scenario-writer subagent to actually write the test in a separate context. When it finishes please validate its work and launch a new subagent with a new task if the user proposed multiple scenarios (or to correct the work of the last one).
Test Structure
Each test in the tests/ directory should have:
tests/
{test-name}/
Reference.svelte - Reference implementation (known-good solution)
test.ts - Vitest test file (imports "./Component.svelte")
prompt.md - Prompt for the AI agent
validator.ts - Extra validation to verify with static analysis that the code written adhere to Svelte best practices
The benchmark:
- Reads the prompt from
prompt.md
- Asks the agent to generate a component
- Writes the generated component to a temporary location
- Runs the tests against the generated component
- Reports pass/fail status
Verifying Reference Implementations
To verify that all reference implementations pass their tests:
bun run verify-tests
This copies each Reference.svelte to Component.svelte temporarily and runs the tests.
