# Getting Started Screenslop is easiest to understand as a loop: ```text capture real UI -> critique evidence -> fix selected issue -> recapture -> verify ``` If you skip the recapture, you do not have a verified fix. You have a hopeful edit. Hope is lovely, but it is not a test. ## 1. Check the machine ```bash node bin/screenslop.mjs doctor ``` The preferred stack is: 1. Baguette 2. XcodeBuildMCP 3. xcodebuild + simctl 4. manual evidence Baguette is the best path for current iOS simulator evidence because it can capture the screenshot, AX tree, and logs from the live surface. If you are checking several booted simulators, Baguette also has a device farm at `http://localhost:8421/farm`. See `docs/baguette-farm.md` for the boundary: farm is useful for observation, while Screenslop bundles remain the proof. ## 2. Create project config Start with setup dry-run from the app repo. It detects the Xcode project, scheme, bundle ID, and source root when there is one safe answer: ```bash node bin/screenslop.mjs setup --json --dry-run ``` If the dry run says `status: "ready"`, write the local config when you are ready: ```bash node bin/screenslop.mjs setup --json --yes ``` If setup says `status: "needs-selection"`, pass the target fields explicitly and dry-run again: ```bash node bin/screenslop.mjs setup \ --project MyApp.xcodeproj \ --scheme MyApp \ --bundle-id com.example.MyApp \ --source-root MyApp \ --surface Settings \ --json \ --dry-run ``` `.screenslop/config.json` is ignored by git because it can contain private paths and bundle IDs. Setup is configuration only; proof starts with runtime capture. Before you run a private app from the Screenslop checkout, check the config shape without launching anything. Pass an absolute config path, or a path relative to this checkout: ```bash node scripts/smoke-real-runtime.mjs \ --config /path/to/private-app/.screenslop/config.json \ --identifier settings.saveButton \ --preflight-only ``` That prints JSON, validates the configured target fields, redacts private paths and bundle IDs, and stops before Baguette, XcodeBuildMCP, build/run, capture, or verify. If this fails, fix the config first. No need to wake the simulator just to discover you misspelled a scheme. Not that any of us would ever do that. ## 3. Capture evidence ```bash node bin/screenslop.mjs see --surface Settings --json ``` The command writes an evidence bundle under `artifacts/` or your configured `artifactsDir`. ## 4. Critique the bundle ```bash node bin/screenslop.mjs critique artifacts/ --json ``` Findings must point back to evidence. If evidence is weak, the finding says so. ## 5. Fix one selected finding ```bash node bin/screenslop.mjs fix artifacts/ \ --finding \ --source-root MyApp \ --apply \ --yes \ --label "Save settings" \ --json ``` The MVP only applies narrow, high-confidence SwiftUI patches. Ambiguous findings become manual plan items instead of surprise edits, which is exactly how it should be. ## 6. Recapture and verify ```bash node bin/screenslop.mjs see --surface Settings --json node bin/screenslop.mjs critique artifacts/ --json node bin/screenslop.mjs verify artifacts/ \ --fresh-bundle artifacts/ \ --finding \ --fix-session artifacts//fix-session.json \ --json ``` `fix-session.json` is context. Fresh evidence is proof. ## Sample app smoke For a self-contained live runtime check: ```bash npm run smoke:runtime ``` This builds `examples/runtime-smoke-app`, captures baseline evidence, applies one safe fix, captures fresh evidence, and verifies the selected finding. It proves the Screenslop loop against the sample app, not against a private app you have not captured yet.