Evenfire is an intelligence layer platform for running LLM agents that take real actions, on your infrastructure
TypeScript
import { describe, expect, it } from 'vitest'
import {
looksLikeWorkflowAccessRequest,
looksLikeWorkflowTriggerRequest,
workflowAccessDeniedResponse,
} from '../providerWorkflowAccessGate'
describe('provider workflow access gate', () => {
it.each([
'List the workflow recipes I can run.',
'what workflows are available?',
'Show workflow details for risk-review',
'Run risk-review with marker alpha',
'Trigger the workflow recipe named risk-review',
'Get the workflow status for risk-review',
'Download the workflow result artifact for risk-review',
'Download artifact for workflow risk-review',
'/approve risk-review',
])('classifies workflow access request: %s', content => {
expect(looksLikeWorkflowAccessRequest(content)).toBe(true)
})
it.each([
'what is a workflow?',
'Explain workflows',
'Define a workflow',
'Hello, summarize this conversation',
'Run a shell command in the workspace',
'Download the generated artifact',
'Send me the output file',
])('allows normal chat or generic explanation: %s', content => {
expect(looksLikeWorkflowAccessRequest(content)).toBe(false)
})
it('detects explicit or workflow-shaped trigger requests without matching generic run text', () => {
expect(looksLikeWorkflowTriggerRequest('Run risk-review with marker alpha')).toBe(true)
expect(looksLikeWorkflowTriggerRequest('Execute the workflow recipe named payroll')).toBe(true)
expect(looksLikeWorkflowTriggerRequest('Run a command')).toBe(false)
expect(
looksLikeWorkflowTriggerRequest(
'List the workflow recipes I can run. Include exact workflow recipe names only.'
)
).toBe(false)
})
it('uses channel-specific verification copy', () => {
expect(workflowAccessDeniedResponse('telegram')).toContain('Telegram conversation')
expect(workflowAccessDeniedResponse('slack')).toContain('Slack workspace conversation')
})
})