@sentinele2e/cli

The unified sentinel command for Sentinel — the operator entry point. It scaffolds a project, runs your flows, and turns the emitted telemetry into a defect classification. The CLI is driver-agnostic: it depends only on @sentinele2e/ai + @sentinele2e/core/@sentinele2e/contracts and shells out to your project's test runner (npx playwright test), so it never imports a browser driver — your project owns that.

Install

npm install -D @sentinele2e/cli

Commands

sentinel init [dir] [--force]            # scaffold a starter project
sentinel run [pattern] [--config <p>] [--dry-run]
                                         # shell out to `npx playwright test`
sentinel analyze <run.jsonl> [--json]    # classify one run (exit 1 iff a real bug)
sentinel report [dir] [--json]           # aggregate a telemetry dir -> RunReport
sentinel --help | --version

• analyze delegates to @sentinele2e/ai (rules-only, no API key) and exits 1 iff a real-bug verdict is present.
• report reads every *.jsonl in dir (default config.telemetryDir), classifies each run, and prints a cross-run table (or --json). Exits 1 iff any run has a real bug; an empty/missing dir is not an error.
• run builds npx playwright test -c <playwrightConfig> [pattern]. --dry-run prints the command and exits 0 without spawning (the offline-testable path). A real run spawns the runner inheriting stdio, then hints at sentinel report. A real spawn needs a project + installed browsers, so it is exercised manually, not in offline CI (spec §5).

Configuration — sentinel.config.json

JSON so the standalone node dist/cli.js bin can load it without a TypeScript loader. Missing file → defaults; no config is required for analyze/report.

| Field | Default | Used by | | ------------------ | ------------------------ | --------------- | | telemetryDir | test-results/telemetry | report | | testDir | tests | scaffold / docs | | runner | playwright | run | | playwrightConfig | playwright.config.ts | run |

RunReport — the dashboard contract

sentinel report --json emits a RunReport (schemaVersion, generatedFrom, runs[] of RunSummary, and cross-run totals). This JSON shape is the input contract the Sentinel dashboard (slice F) renders; both the CLI table and the dashboard read the same structure. The type is exported from the package barrel.

See the root README for the full Sentinel overview.