agentic-mobile-kit 0.7.17 → 0.7.18

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/bin/cli.js CHANGED
@@ -50,7 +50,7 @@ function main([cmd, ...rest]) {
50
50
  printSummary(init(process.cwd(), { tools: list('tools'), ci: list('ci'), pm: flag('pm') }));
51
51
  });
52
52
  } else if (cmd === 'sync') {
53
- run('sync', () => printSync(sync(process.cwd(), { from: flag('from'), ci: list('ci'), pm: flag('pm') })));
53
+ run('sync', () => { if (printSync(sync(process.cwd(), { from: flag('from'), ci: list('ci'), pm: flag('pm') }))) process.exitCode = 1; });
54
54
  } else if (cmd === 'uninstall') {
55
55
  run('uninstall', () => printUninstall(uninstall(process.cwd(), { from: flag('from'), pm: flag('pm') })));
56
56
  } else if (cmd === 'doctor') {
package/lib/init.js CHANGED
@@ -470,6 +470,34 @@ function merge3(current, base, kitNew, from) {
470
470
  return { text: r.stdout, conflicts: r.status };
471
471
  }
472
472
 
473
+ // AGENTS.md's kit section as the team left it, merged with the kit's new one: a team's edit inside it (the screens it
474
+ // builds for, say) stays. A line the kit fills in with what it finds (the versions, the package manager, the app's
475
+ // checks) is the kit's, whatever it says now: there the old version's line, filled in today, stands in for the team's.
476
+ // The screens are a team's decision, so a line with them counts as an edit. The marker lines are the kit's
477
+ function mergeSection(mine, rawOld, kitOld, kitNew, from) {
478
+ const lines = s => s.replace(/\n$/, '').split('\n');
479
+ const [raw, base, theirs] = [rawOld, kitOld, kitNew].map(lines);
480
+ const found = raw.length !== base.length ? [] : raw.map((l, i) => /\{\{\w+\}\}/.test(l) && !l.includes('{{SCREENS}}')
481
+ && [new RegExp(`^${l.split(/\{\{\w+\}\}/).map(s => s.replace(/[.*+?^${}()|[\]\\]/g, '\\$&')).join('.*')}$`), base[i]]).filter(Boolean);
482
+ const ours = lines(mine).map(l => (base.includes(l) ? l : (found.find(([re]) => re.test(l)) || [null, l])[1]));
483
+ const inner = a => `${a.slice(1, -1).join('\n')}\n`;
484
+ const { text, conflicts } = merge3(inner(ours), inner(base), inner(theirs), from);
485
+ return { text: !conflicts && text === inner(theirs) ? kitNew : `${theirs[0]}\n${text}${theirs[theirs.length - 1]}\n`, conflicts };
486
+ }
487
+
488
+ // whether a merged script still parses (two edits can merge cleanly into one that doesn't): sh -n for shell scripts
489
+ // and git hooks, node --check for JavaScript. Without sh on the PATH (Windows without Git's), shell isn't checked
490
+ function parses(rel, text) {
491
+ const shell = rel.endsWith('.sh') || rel.startsWith('.githooks/');
492
+ if (!shell && !/\.[cm]?js$/.test(rel)) return true;
493
+ const dir = fs.mkdtempSync(path.join(os.tmpdir(), 'amk-parse-'));
494
+ const file = path.join(dir, path.basename(rel));
495
+ fs.writeFileSync(file, text);
496
+ const r = spawnSync(shell ? 'sh' : process.execPath, [shell ? '-n' : '--check', file], { encoding: 'utf8' });
497
+ fs.rmSync(dir, { recursive: true, force: true });
498
+ return Boolean(r.error) || r.status === 0;
499
+ }
500
+
473
501
  // Git hook managers a project may already use. The kit's hooks then run from theirs and core.hooksPath stays theirs:
474
502
  // two tools setting it undo each other on every install (husky's prepare script runs after the kit's postinstall).
475
503
  const KIT_HOOKS = ['pre-commit', 'pre-push'];
@@ -489,7 +517,10 @@ const forwarder = h => `#!/bin/sh\n${FORWARD} .git/hooks/${h}, which core.hooksP
489
517
  const localHooksPath = cwd => {
490
518
  let v = '';
491
519
  try { v = git(cwd, 'config', '--local', 'core.hooksPath'); } catch {}
492
- return v && path.posix.normalize(v.replace(/\\/g, '/')).replace(/\/+$/, '');
520
+ v = v && path.posix.normalize(v.replace(/\\/g, '/')).replace(/\/+$/, '');
521
+ // the project's own .githooks as a full path (git config core.hooksPath "$PWD/.githooks") is the kit's folder too
522
+ const real = p => { try { return fs.realpathSync(p); } catch { return path.resolve(p); } };
523
+ return v && path.isAbsolute(v) && real(v) === real(path.join(cwd, '.githooks')) ? '.githooks' : v;
493
524
  };
494
525
 
495
526
  function hookManager(cwd, pkg) {
@@ -662,10 +693,11 @@ function finish(cwd, { pkg, deps, tools, files, stack, skipGenerate }) {
662
693
  const gi = path.join(cwd, '.gitignore');
663
694
  const giText = readIf(gi) || '';
664
695
  // Expo: ios/ and android/ are generated by prebuild, so they stay out of git too
665
- const ignored = ['.env', '.env.*', '!.env*.example', '!.env*.sample', '!.env*.template', '.amk/', ...(stack === 'expo' ? ['/ios', '/android'] : [])];
696
+ // .maestro/.amk-*: the copies of flows with view-tree steps that verify.sh runs on Android, there only while they run
697
+ const ignored = ['.env', '.env.*', '!.env*.example', '!.env*.sample', '!.env*.template', '.amk/', '.maestro/.amk-*', ...(stack === 'expo' ? ['/ios', '/android'] : [])];
666
698
  const missing = ignored.filter(l => !giText.split('\n').includes(l));
667
699
  if (missing.length) {
668
- fs.writeFileSync(gi, `${giText.replace(/\n*$/, giText ? '\n' : '')}\n# agentic-mobile-kit: secrets stay out of git; .amk/ holds local evidence (screenshots, logs)${stack === 'expo' ? '; ios/ and android/ are generated' : ''}\n${missing.join('\n')}\n`);
700
+ fs.writeFileSync(gi, `${giText.replace(/\n*$/, giText ? '\n' : '')}\n# agentic-mobile-kit: secrets stay out of git; .amk/ holds local evidence (screenshots, logs), .maestro/.amk-* the flows verify.sh is running${stack === 'expo' ? '; ios/ and android/ are generated' : ''}\n${missing.join('\n')}\n`);
669
701
  done.merged.push('.gitignore');
670
702
  }
671
703
 
@@ -834,7 +866,7 @@ function sync(cwd, opts = {}) {
834
866
  const oldValues = { ...values, ...ciTexts(hadCi) };
835
867
  const tools = installedTools(cwd);
836
868
  const targets = tools.flatMap(t => TOOLS[t]);
837
- const report = { moved: [], updated: [], added: [], merged: [], conflicts: [], removed: [], kept: [] };
869
+ const report = { moved: [], updated: [], added: [], merged: [], conflicts: [], byHand: [], removed: [], kept: [] };
838
870
  const newFiles = stackFiles(opts.template || TEMPLATE, stack, deps);
839
871
  // an install from before 0.7.0: the kit's files move to the new folders first, so the team's edits come along.
840
872
  // Other files in the old folders may be the app's own, so they stay where they are.
@@ -862,8 +894,14 @@ function sync(cwd, opts = {}) {
862
894
  const kitOld = oldFiles.has(rel) && wasInstalled(rel) ? kitText(oldFiles.get(rel), rel, oldValues, targets) : null;
863
895
  const current = projectText(cwd, rel, targets);
864
896
  if (rel === 'AGENTS.md' || rel === 'CLAUDE.md') {
865
- const next = current === null ? kitNew : rel === 'AGENTS.md' ? mergeAgents(current, kitNew) : mergeClaude(current);
866
- if (next !== current) { fs.writeFileSync(path.join(cwd, rel), next); report.updated.push(rel); }
897
+ // the team's edits inside the kit's section stay, merged like any file's
898
+ const mine = rel === 'AGENTS.md' && exact && kitOld !== null && current !== null && (current.match(MARKERS) || [])[0];
899
+ const { text: section, conflicts } = mine ? mergeSection(mine, fs.readFileSync(oldFiles.get(rel), 'utf8'), kitOld, kitNew, from) : { text: kitNew, conflicts: 0 };
900
+ const next = current === null ? kitNew : rel === 'AGENTS.md' ? mergeAgents(current, section) : mergeClaude(current);
901
+ if (next === current) continue;
902
+ fs.writeFileSync(path.join(cwd, rel), next);
903
+ if (section === kitNew) report.updated.push(rel);
904
+ else (conflicts ? report.conflicts : report.merged).push(`${rel} (your edits inside the kit's section)`);
867
905
  continue;
868
906
  }
869
907
  if (current === null) {
@@ -879,6 +917,14 @@ function sync(cwd, opts = {}) {
879
917
  if (kitOld === kitNew || PROJECT_DOCS.includes(rel)) continue;
880
918
  // without an exact base, an old unchanged file would look like your edit and be kept: show every difference instead
881
919
  const { text, conflicts } = merge3(current, exact ? kitOld : '', kitNew, from);
920
+ // markers, or a merge that no longer parses, would stop a script, a hook, a workflow or a config from working (and
921
+ // every check with it): those get the kit's version, and the team's goes beside it to bring its changes back
922
+ if (!rel.endsWith('.md') && (conflicts || !parses(rel, text))) {
923
+ fs.copyFileSync(path.join(cwd, rel), path.join(cwd, `${rel}.mine`));
924
+ writeKitFile(cwd, rel, kitNew);
925
+ report.byHand.push(rel);
926
+ continue;
927
+ }
882
928
  writeKitFile(cwd, rel, text);
883
929
  (conflicts ? report.conflicts : report.merged).push(rel);
884
930
  }
@@ -1014,13 +1060,19 @@ function printSync({ report, warnings = [], from, to, changed, generated, genera
1014
1060
  log(from === to ? `agentic-mobile-kit is already ${to}.` : `agentic-mobile-kit updated from ${from} to ${to}.`);
1015
1061
  const lines = [
1016
1062
  ['moved to the new folders', report.moved || []], ['updated', report.updated], ['added', report.added], ['removed', report.removed],
1017
- ['merged with your changes', report.merged], ['CONFLICTS, resolve the <<<<<<< markers', report.conflicts], ['left alone', report.kept],
1063
+ ['merged with your changes', report.merged], ['CONFLICTS, resolve the <<<<<<< markers', report.conflicts],
1064
+ ['CONFLICTS where markers would stop the file from working, so the kit\'s version is in place: bring your changes back from <file>.mine, then delete it', report.byHand || []],
1065
+ ['left alone', report.kept],
1018
1066
  ];
1019
1067
  for (const [label, files] of lines) if (files.length) log(` ${label}: ${files.join(', ')}`);
1020
1068
  if (!changed) log(' Nothing changed.');
1021
1069
  for (const w of warnings) log(` ! ${w}`);
1022
1070
  if (!generated) log(` ! could not generate tool configs; run: npx -y ${RULESYNC} generate\n${generateError.replace(/^/gm, ' ')}`);
1023
- if (changed) log('\nReview it with "git diff", run the checks, then commit on a branch and open a PR.');
1071
+ // a conflict left is an update not finished: said last, and the command fails
1072
+ const open = report.conflicts.length + (report.byHand || []).length;
1073
+ if (open) log(`\nsync isn't finished: ${open === 1 ? 'a conflict' : `${open} conflicts`} to resolve (above). Then run the checks, and commit.`);
1074
+ else if (changed) log('\nReview it with "git diff", run the checks, then commit on a branch and open a PR.');
1075
+ return open;
1024
1076
  }
1025
1077
 
1026
1078
  function printUninstall({ report }, log = console.log) {
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "agentic-mobile-kit",
3
- "version": "0.7.17",
3
+ "version": "0.7.18",
4
4
  "description": "One setup for AI-assisted development of React Native apps, bare or Expo: shared rules and workflows for Claude Code, Codex and Antigravity, git hooks, device checks with screenshots, and release tooling for the App Store and Google Play.",
5
5
  "license": "MIT",
6
6
  "author": "prakashpro1",
@@ -20,9 +20,9 @@ This project may also use other AI workflow frameworks (for example Superpowers,
20
20
  | Add a library (ask first) | `npx expo install <package>`: the version that matches the SDK |
21
21
  | Lint, typecheck, test (each skips what the app doesn't have yet) | `sh scripts/amk/pm-run.sh lint`, `sh scripts/amk/pm-run.sh typecheck`, `sh scripts/amk/pm-run.sh test` |
22
22
  | Check dependency versions and the app config | `npx expo install --check`, `npx expo-doctor` |
23
- | Check on devices (release builds, Maestro flows, screenshots) | `sh scripts/amk/verify.sh all --spec <id>`; another iOS scheme: `--scheme <name>`; only some screens: `--screens phone` (or `tablet`, `landscape`, …); on a real iPhone or iPad the user names: `AMK_IOS_DEVICE=<its UDID>` (the flows clear the app's data on it) |
23
+ | Check on devices (release builds, Maestro flows, screenshots) | `sh scripts/amk/verify.sh all --spec <id>`; another iOS scheme: `--scheme <name>`; only some screens: `--screens phone` (or `tablet`, `landscape`, …); only the flows tagged smoke: `--tags smoke`; on a real iPhone or iPad the user names: `AMK_IOS_DEVICE=<its UDID>` (the flows clear the app's data on it) |
24
24
  | Check devices at the same time | `sh scripts/amk/verify.sh plan --spec <id>` says how many this machine's memory fits, and prints the commands: one per device (in rounds when they don't all fit; a device's flows in parts when there's room), each for an `m-device-checker` subagent |
25
- | React Native mistakes in this branch's changed lines (`verify.sh quick` runs it) | `node scripts/amk/rn-check.js` |
25
+ | React Native mistakes and Maestro flow problems in this branch's changed lines (`verify.sh quick` runs it) | `node scripts/amk/rn-check.js` |
26
26
  | Specs: progress, format check, fold a finished feature into the living spec | `node scripts/amk/specs.js status`, `node scripts/amk/specs.js check`, `node scripts/amk/specs.js merge <id>` |
27
27
  | Check a release before tagging it (versions, permission texts, store review rules, debug leftovers) | `node scripts/amk/release-check.js` |
28
28
  | Signed release builds (a person runs them: they use the signing keys) | `sh scripts/amk/release-build.sh android`, `sh scripts/amk/release-build.sh ios` |
@@ -40,6 +40,7 @@ Never say a check passed unless you ran it in this session.
40
40
  ## Rules
41
41
 
42
42
  - Stay inside the task. Ask before touching unrelated files.
43
+ - Before designing a feature's screens, ask the user which UI/UX level they want: Simple (the default), Mixed or Premium (`specs/_templates/design.md` describes each). Build them to it.
43
44
  - Ask before adding or upgrading dependencies, and before editing `package.json`, lockfiles, `app.json`, `app.config.*`, `eas.json` or `.github/workflows/`.
44
45
  - Never edit `ios/` or `android/`: the next prebuild overwrites them. Native settings go in `app.json` or `app.config.*`, through config plugins; native code goes in an Expo module (`expo-module` skill). Ask before adding a config plugin.
45
46
  - Never run `eas build`, `eas submit` or `eas update`: they build, ship or publish with the team's accounts, so a person runs them.
@@ -35,17 +35,18 @@ If the input describes a bug rather than new behavior, suggest `m-bugfix` and st
35
35
 
36
36
  ## 3. Design and tasks (then stop for approval)
37
37
 
38
- 1. **Search the codebase before designing.** List the components, hooks, API clients and helpers you'll reuse in `design.md` under "Existing code to reuse". Read `docs/amk/structure.md` and `docs/amk/conventions.md`.
39
- 2. Fill `design.md`: approach, files to change, risks (native changes, pod install, migrations).
40
- 3. Fill `tasks.md`: small tasks (about an hour, under ~300 changed lines each). Each task names the requirement it serves and how it's checked, and its tests are written in the same task, not left for the end. The last task writes `.maestro/<id>.yaml` (from `specs/_templates/flow.yaml`): a Maestro flow with one step and screenshot per scenario, tapping elements by `testID` and scrolling to the ones further down a screen first (`scrollUntilVisible`): turned screens are short.
41
- 4. Show design and tasks and **stop until the user approves them**. Ask them to check that every task serves an approved requirement and nothing goes beyond the agreed scope.
38
+ 1. **Ask for the UI/UX level first, when the feature has screens.** As one question, ask which level its screens should have: Simple (clean, minimal and functional), Mixed (simple, with modern touches) or Premium (polished, following current design trends; it takes longer). `specs/_templates/design.md` describes each. Offer as the default the level a prototype settled (under `## Decisions` in `requirements.md`), otherwise the one the app's other screens have (an earlier spec's `design.md`), otherwise Simple. Don't design until the user has answered. Write the answer and the date under "UI/UX level" in `design.md` (n/a for a feature without screens).
39
+ 2. **Search the codebase before designing.** List the components, hooks, API clients and helpers you'll reuse in `design.md` under "Existing code to reuse". Read `docs/amk/structure.md` and `docs/amk/conventions.md`.
40
+ 3. Fill `design.md`: approach, files to change, risks (native changes, pod install, migrations).
41
+ 4. Fill `tasks.md`: small tasks (about an hour, under ~300 changed lines each). Each task names the requirement it serves and how it's checked, and its tests are written in the same task, not left for the end. The last task writes `.maestro/<id>.yaml` (from `specs/_templates/flow.yaml`): a Maestro flow with a `# Scenario <n>: <name>` comment above each scenario's steps, an assertion for its THEN and a screenshot, tapping elements by `testID` and scrolling to the ones further down a screen first (`scrollUntilVisible`): turned screens are short.
42
+ 5. Show design and tasks and **stop until the user approves them**. Ask them to check that every task serves an approved requirement and nothing goes beyond the agreed scope.
42
43
 
43
44
  From here the plan lives in files, so the building can also run in a fresh session or another tool with `m-continue <id>`.
44
45
 
45
46
  ## 4. Build, one task at a time
46
47
 
47
48
  For each task:
48
- 1. Implement only that task. Write tests first for logic (state, data, utils); UI can be tested after. Build screens to the screen rules in the React Native rules file `AGENTS.md` names (keyboard, system bars, scrolling, large text, announced messages), for every phone, tablet and orientation `AGENTS.md` lists: they're the bugs device checks and reviews find most.
49
+ 1. Implement only that task. Write tests first for logic (state, data, utils); UI can be tested after. Build screens to the UI/UX level in `design.md`, and to the screen rules in the React Native rules file `AGENTS.md` names (keyboard, system bars, scrolling, large text, announced messages), for every phone, tablet and orientation `AGENTS.md` lists: they're the bugs device checks and reviews find most.
49
50
  2. Run `lint`, `typecheck` and `test` (see AGENTS.md). Paste the real output. If something fails, fix it; after 2 failed attempts, stop and report.
50
51
  3. Commit with a clear message ending in `Spec: specs/<id>`.
51
52
  4. Tick the task in `tasks.md` and update the handoff block and log in `progress.md`.
@@ -16,9 +16,10 @@ A prototype answers questions before the real build: what the client wants, how
16
16
 
17
17
  1. Read the requirement, `docs/amk/product.md`, `structure.md` and `conventions.md`, the living spec for the area in `specs/current/`, and the screens around where this would live: their components, theme and navigation.
18
18
  2. Ask who the demo is for (the client or the team) and what it must show, one question at a time.
19
- 3. Pick an ID: the task's ID if it has one, otherwise `YYYYMMDD-<slug>`. Create the branch `prototype/<id>` from the main branch; this workflow commits everything there.
20
- 4. Write `specs/<id>/requirements.md` from `specs/_templates/requirements.md`: the goal, what's known so far, and the open questions the prototype should answer. Add a `## Decisions` section for what gets settled along the way.
21
- 5. Decide which case this is:
19
+ 3. Ask which UI/UX level the screens should have: Simple (clean, minimal and functional; the default), Mixed (simple, with modern touches) or Premium (polished, following current design trends). `specs/_templates/design.md` describes each; offer the level the app's other screens have, if an earlier spec's `design.md` records one. The concepts and the prototype are made at that level. Don't start them until the user has answered.
20
+ 4. Pick an ID: the task's ID if it has one, otherwise `YYYYMMDD-<slug>`. Create the branch `prototype/<id>` from the main branch; this workflow commits everything there.
21
+ 5. Write `specs/<id>/requirements.md` from `specs/_templates/requirements.md`: the goal, what's known so far, and the open questions the prototype should answer. Add a `## Decisions` section for what gets settled along the way, starting with the UI/UX level and the date.
22
+ 6. Decide which case this is:
22
23
  - **What to show is clear** (the screens and the flow are known): go to step 3.
23
24
  - **How it should work or look is still open:** go to step 2.
24
25
 
@@ -17,7 +17,7 @@ If your tool has subagents, hand the review to the `m-reviewer` subagent, so it
17
17
  - **Against the spec**, when there is one:
18
18
  - *Complete:* every task is ticked, every requirement is implemented, and every scenario has evidence (a test or a screenshot).
19
19
  - *Correct:* the code does what each requirement says, edge cases included, and nothing outside the agreed scope.
20
- - *Coherent:* the decisions in `design.md` show up in the code, and `specs/current/` was updated (`node scripts/amk/specs.js merge <id>`) if behavior changed.
20
+ - *Coherent:* the decisions in `design.md` show up in the code, its UI/UX level in the screens, and `specs/current/` was updated (`node scripts/amk/specs.js merge <id>`) if behavior changed.
21
21
  - **No spec:** does the diff do what was asked, and nothing unrelated?
22
22
  - **Tests:** were tests added for new logic? Were any tests deleted, skipped or weakened?
23
23
  - **TypeScript:** any new `any`, `@ts-ignore` or non-null `!` to get past a check, or a loosened `tsconfig.json`?
@@ -65,6 +65,9 @@ jobs:
65
65
  disable-animations: true
66
66
  emulator-options: -no-window -gpu swiftshader_indirect -noaudio -no-boot-anim -camera-back none
67
67
  script: sh scripts/amk/verify.sh android --label ci
68
+ - name: Summary (each flow's result, from verify.sh's report)
69
+ if: always()
70
+ run: if [ -d .amk/evidence/ci ]; then node scripts/amk/report.js .amk/evidence/ci --summary >> "$GITHUB_STEP_SUMMARY"; fi
68
71
  - name: Upload screenshots and logs
69
72
  if: always()
70
73
  uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9 # v7.0.2
@@ -9,8 +9,8 @@ claudecode:
9
9
  You check the app on one device with fresh eyes, and hand the evidence back. You never edit files.
10
10
 
11
11
  1. Run the one command you were given exactly as it is, for example `sh scripts/amk/verify.sh ios --no-build --screens tablet --spec <id>`: no pipes, redirects or other commands added. Its output is short, and the plain command is the one people allow, so you don't wait on a permission prompt while the other checkers run. It installs the build that's there, runs the feature's Maestro flows on that device's screens (upright and turned), and saves screenshots in `.amk/evidence/<id>/<screen>/`. A command with `--part k/n` runs only its share of the flows, on a device of its own (a temporary copy for the parts after the first, deleted afterwards), and saves them in `.amk/evidence/<id>/<screen>/part-k/`. Don't build, and don't run anything else on a device: other checkers are using the other devices. A check takes several minutes, up to about 15 when a flow has to run again: give the command a timeout that long, or run it in the background and wait for it to end, never stopping it early.
12
- 2. If a flow failed, read its log and Maestro's own log (the output names both) and say why, and whether the app or the flow is wrong: on a turned screen, a flow often checks or taps something further down without scrolling to it first. Don't fix anything.
13
- 3. Open every screenshot in your screens' folders and check each against its scenario in `specs/<id>/requirements.md`, and against the screen rules in the React Native rules file that `AGENTS.md` names: the right content, nothing cut off or covered, a layout made for that screen. A turned screen's screenshots are landscape; a portrait one means the app didn't turn.
12
+ 2. If a flow failed, the output says at which step and what Maestro said, and shows the screen then (`report.md` in the evidence folder has it too); read its log and Maestro's own log for more, and say why, and whether the app or the flow is wrong: on a turned screen, a flow often checks or taps something further down without scrolling to it first. Don't fix anything.
13
+ 3. On Android screens, the output may list accessibility warnings from the screens' view trees (controls without a label, tap areas under 48 dp; `report.md` says which screenshot): report them as they are, they don't fail the check. Open every screenshot in your screens' folders and check each against its scenario in `specs/<id>/requirements.md`, and against the screen rules in the React Native rules file that `AGENTS.md` names: the right content, nothing cut off or covered, a layout made for that screen. A turned screen's screenshots are landscape; a portrait one means the app didn't turn.
14
14
  4. Report, and say only what you saw in this run:
15
15
  - the command, and its last lines of output;
16
16
  - per screen: flows passed out of total, and the evidence folder;
@@ -20,9 +20,9 @@ This project may also use other AI workflow frameworks (for example Superpowers,
20
20
  | iOS pods (after native dependency changes) | `sh scripts/amk/pod-install.sh` |
21
21
  | iOS build (simulator, no signing) | `sh scripts/amk/ios-build.sh` |
22
22
  | Android build | `cd android && ./gradlew assembleDebug` |
23
- | Check on devices (release builds, Maestro flows, screenshots) | `sh scripts/amk/verify.sh all --spec <id>`; with product flavors it checks the first unless `--flavor <name>`; another iOS scheme: `--scheme <name>`; only some screens: `--screens phone` (or `tablet`, `landscape`, …); on a real iPhone or iPad the user names: `AMK_IOS_DEVICE=<its UDID>` (the flows clear the app's data on it) |
23
+ | Check on devices (release builds, Maestro flows, screenshots) | `sh scripts/amk/verify.sh all --spec <id>`; with product flavors it checks the first unless `--flavor <name>`; another iOS scheme: `--scheme <name>`; only some screens: `--screens phone` (or `tablet`, `landscape`, …); only the flows tagged smoke: `--tags smoke`; on a real iPhone or iPad the user names: `AMK_IOS_DEVICE=<its UDID>` (the flows clear the app's data on it) |
24
24
  | Check devices at the same time | `sh scripts/amk/verify.sh plan --spec <id>` says how many this machine's memory fits, and prints the commands: one per device (in rounds when they don't all fit; a device's flows in parts when there's room), each for an `m-device-checker` subagent |
25
- | React Native mistakes in this branch's changed lines (`verify.sh quick` runs it) | `node scripts/amk/rn-check.js` |
25
+ | React Native mistakes and Maestro flow problems in this branch's changed lines (`verify.sh quick` runs it) | `node scripts/amk/rn-check.js` |
26
26
  | Specs: progress, format check, fold a finished feature into the living spec | `node scripts/amk/specs.js status`, `node scripts/amk/specs.js check`, `node scripts/amk/specs.js merge <id>` |
27
27
  | Check a release before tagging it (versions, permission texts, store review rules, debug leftovers) | `node scripts/amk/release-check.js` |
28
28
  | Signed release builds (a person runs them: they use the signing keys) | `sh scripts/amk/release-build.sh android [flavor]`, `sh scripts/amk/release-build.sh ios [scheme]` |
@@ -40,6 +40,7 @@ Never say a check passed unless you ran it in this session.
40
40
  ## Rules
41
41
 
42
42
  - Stay inside the task. Ask before touching unrelated files.
43
+ - Before designing a feature's screens, ask the user which UI/UX level they want: Simple (the default), Mixed or Premium (`specs/_templates/design.md` describes each). Build them to it.
43
44
  - Ask before adding or upgrading dependencies, and before editing `package.json`, lockfiles, `Podfile`, `build.gradle`, `Info.plist`, `AndroidManifest.xml` or `.github/workflows/`.
44
45
  - Don't hand-edit `ios/*.xcodeproj/project.pbxproj`. If a native file must be added to the Xcode project, stop and ask.
45
46
  - Never read, print or commit `.env` files, keystores, signing files or the files `.amk-secrets` lists. Secrets don't belong in this repo.
@@ -0,0 +1,78 @@
1
+ #!/usr/bin/env node
2
+ // Accessibility on the device: from the view trees verify.sh has Maestro save after each screenshot on Android (which
3
+ // marks what's tappable; iOS' tree doesn't), the tappable elements without a label, which a screen reader can't name,
4
+ // and the ones smaller than 48×48 dp, Android's minimum (and more than iOS' 44 points: React Native lays out both
5
+ // alike). An element a scrolling list cuts off at its edge isn't measured. Prints a line each, once per element and
6
+ // kind of problem in a flow: flow, screenshot, problem, element.
7
+ // Usage: node scripts/amk/a11y.js <Maestro debug folder> <screen density in dpi> <flow>
8
+ 'use strict';
9
+ const fs = require('fs');
10
+ const path = require('path');
11
+
12
+ const [dir, dpi, flow] = process.argv.slice(2);
13
+ const scale = Number(dpi) / 160;
14
+ if (!dir || !scale || !fs.existsSync(dir)) process.exit(0);
15
+
16
+ // the trees and the screenshot each follows, in the order the flow saved them. Maestro's commands.json has each step:
17
+ // the view-tree step's id (amk-view-tree-<screenshot>) and the tree it saved; the file's own name is cut short
18
+ const trees = [];
19
+ const files = [];
20
+ (function find(d, depth) {
21
+ for (const e of fs.readdirSync(d, { withFileTypes: true })) {
22
+ const p = path.join(d, e.name);
23
+ if (e.isDirectory() && depth < 4) find(p, depth + 1);
24
+ else files.push(p);
25
+ }
26
+ })(dir, 0);
27
+ for (const f of files.filter(p => path.basename(p) === 'commands.json')) {
28
+ let steps;
29
+ try { steps = JSON.parse(fs.readFileSync(f, 'utf8')); } catch { continue; }
30
+ for (const s of Array.isArray(steps) ? steps : []) {
31
+ const id = ((((s.command || {}).assertConditionCommand || {}).condition || {}).visible || {}).idRegex || '';
32
+ const tree = ((s.metadata || {}).artifacts || []).find(a => a.type === 'SCREEN_HIERARCHY');
33
+ if (id.startsWith('amk-view-tree-') && tree) trees.push([path.join(path.dirname(f), tree.path), id.slice(14)]);
34
+ }
35
+ }
36
+ // without commands.json: the names of the trees' files
37
+ if (!trees.length) {
38
+ for (const p of files.filter(p => /amk-view-tree-.+\.json$/.test(p)).sort((a, b) => path.basename(a).localeCompare(path.basename(b)))) {
39
+ trees.push([p, /amk-view-tree-(.+)\.json$/.exec(path.basename(p))[1]]);
40
+ }
41
+ }
42
+
43
+ // the system's own windows, with everything inside them (the status bar, a keyboard and its suggestions); Android's
44
+ // own ids (android:id/…) only for that element, as the app's views sit inside android:id/content
45
+ const SYSTEM = /^(com\.android\.systemui|com\.google\.android\.inputmethod|com\.android\.inputmethod|com\.samsung\.android\.honeyboard|com\.touchtype\.swiftkey)/;
46
+ const box = a => { const b = /\[(-?\d+),(-?\d+)\]\[(-?\d+),(-?\d+)\]/.exec(a.bounds || ''); return b && b.slice(1).map(Number); };
47
+ const seen = new Set();
48
+ for (const [file, shot] of trees) {
49
+ let tree;
50
+ try { tree = JSON.parse(fs.readFileSync(file, 'utf8')); } catch { continue; }
51
+ // each node with the bounds of the nearest scrolling container around it
52
+ (function walk(n, clip) {
53
+ if (!n) return;
54
+ const a = n.attributes || {};
55
+ if (SYSTEM.test(a['resource-id'] || '')) return;
56
+ const b = box(a);
57
+ if (a.clickable === 'true' && b && !/^android:id\//.test(a['resource-id'] || '')) {
58
+ const [x1, y1, x2, y2] = b;
59
+ const w = (x2 - x1) / scale;
60
+ const h = (y2 - y1) / scale;
61
+ const label = (a.accessibilityText || a.text || a.hintText || '').trim();
62
+ const id = a['resource-id'] || '';
63
+ const name = id ? `testID ${id}` : label ? `"${label}"` : `${a.class || 'an element'} at ${a.bounds}`;
64
+ // once per element and kind of problem in a flow, where it first shows (its size can differ a little later)
65
+ const say = problem => {
66
+ const key = `${id || label}|${problem.startsWith('no label') ? 'label' : 'size'}`; // a testID names it, whatever it shows
67
+ if (!seen.has(key)) { seen.add(key); console.log([flow, shot, problem, name].join('\t')); }
68
+ };
69
+ if (w > 0 && h > 0) {
70
+ if (!label) say('no label: a screen reader can\'t say what it does');
71
+ const cut = clip && (y1 <= clip[1] || y2 >= clip[3] || x1 <= clip[0] || x2 >= clip[2]);
72
+ if (!cut && (w < 47.5 || h < 47.5)) say(`${Math.round(w)}×${Math.round(h)} dp: under 48×48`);
73
+ }
74
+ }
75
+ const inner = a.scrollable === 'true' && b ? b : clip;
76
+ for (const c of n.children || []) walk(c, inner);
77
+ })(tree.root || tree, null);
78
+ }
@@ -0,0 +1,78 @@
1
+ #!/usr/bin/env node
2
+ // A device check's report, from its evidence folder (.amk/evidence/<label>/): report.md for people (each screen, its
3
+ // flows, passed or failed and why, and their screenshots) and junit.xml for CI. --summary prints a short Markdown
4
+ // table instead, for a CI job's summary. verify.sh runs it after each device check, and builds it from what's on disk,
5
+ // so devices and parts checked at the same time all end up in it: each screen's flows are in <screen>/.flows (or
6
+ // <screen>/part-<k>/.flows), a line each: name, passed or failed, why, its screenshots, the screen when it failed.
7
+ // Usage: node scripts/amk/report.js <evidence folder> [--summary]
8
+ 'use strict';
9
+ const fs = require('fs');
10
+ const path = require('path');
11
+
12
+ const [dir, mode] = process.argv.slice(2);
13
+ if (!dir || !fs.existsSync(dir)) {
14
+ console.error('usage: node scripts/amk/report.js <evidence folder> [--summary]');
15
+ process.exit(2);
16
+ }
17
+ const label = path.basename(path.resolve(dir));
18
+ const read = f => { try { return fs.readFileSync(f, 'utf8'); } catch { return null; } };
19
+
20
+ // the screens: <screen>/.flows, and <screen>/part-<k>/.flows for a device's flows split in parts
21
+ const screens = [];
22
+ for (const s of fs.readdirSync(dir, { withFileTypes: true }).filter(e => e.isDirectory()).map(e => e.name).sort()) {
23
+ for (const sub of ['', ...(fs.readdirSync(path.join(dir, s)).filter(p => /^part-\d+$/.test(p)).sort())]) {
24
+ const text = read(path.join(dir, s, sub, '.flows'));
25
+ if (text === null) continue;
26
+ // what the view trees showed on Android: flow, screenshot, problem, element (a11y.js)
27
+ const a11y = (read(path.join(dir, s, sub, '.a11y')) || '').split('\n').filter(Boolean).map(l => l.split('\t'));
28
+ const flows = text.split('\n').filter(Boolean).map(l => {
29
+ const [name, result, why, shots, shot] = l.split('\t');
30
+ return { name, passed: result === 'passed', why: why || '', shots: (shots || '').split(',').filter(Boolean), shot: shot || '',
31
+ a11y: a11y.filter(x => x[0] === name).map(([, at, problem, element]) => ({ at, problem, element })) };
32
+ });
33
+ screens.push({ name: sub ? `${s} (${sub.replace('-', ' ')})` : s, dir: sub ? `${s}/${sub}` : s, flows });
34
+ }
35
+ }
36
+ const all = screens.flatMap(s => s.flows);
37
+ const failed = all.filter(f => !f.passed);
38
+ const counts = s => `${s.flows.filter(f => f.passed).length}/${s.flows.length}`;
39
+ const table = ['| Screen | Flows passed | Screenshots |', '|---|---|---|',
40
+ ...screens.map(s => `| ${s.name} | ${counts(s)}${s.flows.every(f => f.passed) ? '' : ' ✗'} | ${s.flows.reduce((n, f) => n + f.shots.length, 0)} |`)];
41
+ const warned = all.reduce((n, f) => n + f.a11y.length, 0);
42
+ const headline = `${all.length - failed.length} of ${all.length} flows passed on ${screens.length} screen${screens.length === 1 ? '' : 's'}${warned ? `, ${warned} accessibility warning${warned === 1 ? '' : 's'}` : ''}`;
43
+
44
+ if (mode === '--summary') {
45
+ console.log([`### Device checks: ${label}`, '', headline, '', ...table, '',
46
+ ...failed.map(f => `- ✗ ${screens.find(s => s.flows.includes(f)).name}: ${f.name}${f.why ? `, failed ${f.why}` : ''}`)].join('\n'));
47
+ process.exit(0);
48
+ }
49
+
50
+ // a path inside the evidence folder, for a link from report.md, which sits in it
51
+ const link = p => p.split('/').map(encodeURIComponent).join('/');
52
+ const md = [`# Device checks: ${label}`, '', `${new Date().toISOString().slice(0, 16).replace('T', ' ')} UTC · ${headline}`, '', ...table];
53
+ for (const s of screens) {
54
+ md.push('', `## ${s.name}`);
55
+ for (const f of s.flows) {
56
+ md.push('', `### ${f.passed ? '✓' : '✗'} ${f.name}`);
57
+ if (!f.passed) {
58
+ md.push('', `Failed${f.why ? ` ${f.why}` : ''}. Log: [${s.dir}/${f.name}.log](${link(`${s.dir}/${f.name}.log`)})${f.shot ? ` · [the screen when it failed](${link(f.shot)})` : ''}`);
59
+ }
60
+ if (f.shots.length) md.push('', f.shots.map(p => `<img src="${link(p)}" width="180" alt="${path.basename(p, '.png')}" title="${path.basename(p, '.png')}">`).join(' '));
61
+ if (f.a11y.length) md.push('', 'Accessibility, from the screen\'s view tree:', '', ...f.a11y.map(x => `- ${x.at}: ${x.element}: ${x.problem}`));
62
+ }
63
+ }
64
+ fs.writeFileSync(path.join(dir, 'report.md'), `${md.join('\n')}\n`);
65
+
66
+ const xml = s => String(s).replace(/&/g, '&amp;').replace(/</g, '&lt;').replace(/>/g, '&gt;').replace(/"/g, '&quot;');
67
+ fs.writeFileSync(path.join(dir, 'junit.xml'), [
68
+ '<?xml version="1.0" encoding="UTF-8"?>',
69
+ `<testsuites name="${xml(label)}" tests="${all.length}" failures="${failed.length}">`,
70
+ ...screens.flatMap(s => [
71
+ ` <testsuite name="${xml(s.name)}" tests="${s.flows.length}" failures="${s.flows.filter(f => !f.passed).length}">`,
72
+ ...s.flows.map(f => (f.passed ? ` <testcase classname="${xml(s.name)}" name="${xml(f.name)}"/>`
73
+ : ` <testcase classname="${xml(s.name)}" name="${xml(f.name)}"><failure message="${xml(f.why || 'failed')}"/></testcase>`)),
74
+ ' </testsuite>',
75
+ ]),
76
+ '</testsuites>', '',
77
+ ].join('\n'));
78
+ console.log(`${headline}: ${path.join(dir, 'report.md')}`);
@@ -1,6 +1,8 @@
1
1
  #!/usr/bin/env node
2
2
  // Checks the lines this branch adds or changes for React Native mistakes that the app's rules (docs/amk/react-native.md
3
- // or expo.md) list. Problems (✗) fail it, and so verify.sh quick; warnings (!) don't. Only changed lines: code already
3
+ // or expo.md) list, and its Maestro flows (.maestro/): Maestro's own syntax check, files they run that aren't there,
4
+ // taps by text, and each scenario of a spec the branch touched with its steps, an assertion and a screenshot.
5
+ // Problems (✗) fail it, and so verify.sh quick; warnings (!) don't. Only changed lines: code already
4
6
  // on the base branch is never reported. A line that's right as it is carries `// amk-ok <date>: <reason>`, on it or on
5
7
  // the line above, with the user's reason and the day they gave it; it's listed as kept, for review. When a branch changes
6
8
  // a file again, a kept line in it that's over two weeks old is reported again, for the user to decide again (the date
@@ -32,18 +34,21 @@ const since = baseRef || 'the last commit';
32
34
  const source = f => /\.(js|jsx|ts|tsx|mjs|cjs)$/.test(f) && !f.endsWith('.d.ts')
33
35
  && !/(^|\/)(node_modules|scripts\/amk|__tests__|__mocks__|e2e)\//.test(f)
34
36
  && !/\.(test|spec)\.\w+$|(^|\/)(jest[.-]setup|setup-?tests)[^/]*$/i.test(f);
37
+ // Maestro flows: the ones verify.sh runs (.maestro/*.yaml) and the subflows they run (in folders below)
38
+ // (not the copies with view-tree steps verify.sh runs for a moment on Android: .amk-<screen>-<flow>.yaml)
39
+ const flowFile = f => /^\.maestro\/.+\.ya?ml$/.test(f) && !/(^|\/)\.amk-/.test(f);
35
40
  const changed = new Map(); // file -> Set of line numbers, or 'all'
36
41
  if (sha) {
37
42
  let file = null;
38
43
  for (const line of (git('diff', '-U0', '--no-color', '--no-ext-diff', '--relative', '-M', '--diff-filter=ACMR', sha) || '').split('\n')) {
39
44
  if (line.startsWith('+++ ')) { file = line === '+++ /dev/null' ? null : line.slice(6); continue; }
40
45
  const h = /^@@ -\S+ \+(\d+)(?:,(\d+))? @@/.exec(line);
41
- if (!h || !file || !source(file)) continue;
46
+ if (!h || !file || !(source(file) || flowFile(file))) continue;
42
47
  if (!changed.has(file)) changed.set(file, new Set());
43
48
  for (let n = Number(h[1]), end = n + Number(h[2] ?? 1); n < end; n++) changed.get(file).add(n);
44
49
  }
45
50
  }
46
- for (const f of (git('ls-files', '--others', '--exclude-standard') || '').split('\n')) if (f && source(f)) changed.set(f, 'all');
51
+ for (const f of (git('ls-files', '--others', '--exclude-standard') || '').split('\n')) if (f && (source(f) || flowFile(f))) changed.set(f, 'all');
47
52
 
48
53
  // ---------- what the app uses ----------
49
54
  let pkg = {};
@@ -107,6 +112,13 @@ function okSince(file, n, date) {
107
112
  const t = /^committer-time (\d+)$/m.exec(git('blame', '--porcelain', '-L', `${n},${n}`, '--', file) || '');
108
113
  return t ? localDay(Number(t[1]) * 1000) : today;
109
114
  }
115
+ const BASE64 = new Set([...'ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/='].map(c => c.charCodeAt(0)));
116
+ const SECRETS = [
117
+ [/\bgh[pousr]_[A-Za-z0-9]{36,}/g, 'a GitHub token'], [/\bgithub_pat_[A-Za-z0-9_]{40,}/g, 'a GitHub token'],
118
+ [/\b[sr]k_live_[A-Za-z0-9]{20,}/g, 'a Stripe secret key'], [/\bAKIA[0-9A-Z]{16}\b/g, 'an AWS access key'],
119
+ [/-----BEGIN [A-Z ]*PRIVATE KEY-----/g, 'a private key'], [/\bxox[baprs]-[A-Za-z0-9-]{10,}/g, 'a Slack token'],
120
+ [/\bsk-(proj-|ant-)?[A-Za-z0-9_-]{32,}/g, 'an AI service API key'],
121
+ ];
110
122
  const importsFrom = (code, from) => [...code.matchAll(new RegExp(`import\\s+[^;'"]*?\\{([^}]*)\\}\\s*from\\s*['"]${from}['"]`, 'g'))];
111
123
  const timers = /\bset(Timeout|Interval)\s*\(/;
112
124
  // a permission prompt (not App Tracking Transparency's, which apps show early on purpose)
@@ -155,15 +167,17 @@ function check(file, text) {
155
167
  }
156
168
  });
157
169
  // secrets are reported without their value; comments too, as they ship in the repository
158
- const secrets = [
159
- [/\bgh[pousr]_[A-Za-z0-9]{36,}/g, 'a GitHub token'], [/\bgithub_pat_[A-Za-z0-9_]{40,}/g, 'a GitHub token'],
160
- [/\b[sr]k_live_[A-Za-z0-9]{20,}/g, 'a Stripe secret key'], [/\bAKIA[0-9A-Z]{16}\b/g, 'an AWS access key'],
161
- [/-----BEGIN [A-Z ]*PRIVATE KEY-----/g, 'a private key'], [/\bxox[baprs]-[A-Za-z0-9-]{10,}/g, 'a Slack token'],
162
- [/\bsk-(proj-|ant-)?[A-Za-z0-9_-]{32,}/g, 'an AI service API key'],
163
- ];
164
- for (const [re, what] of secrets) each(re, m => add('✗', m.index, null, `${what} in code: anyone with the repository or the app can read it. Revoke it and keep it on a server`), text);
170
+ for (const [re, what] of SECRETS) each(re, m => add('✗', m.index, null, `${what} in code: anyone with the repository or the app can read it. Revoke it and keep it on a server`), text);
165
171
  each(/\bAIza[0-9A-Za-z_-]{35}\b/g, m => add('!', m.index, null, 'a Google API key in code: ship it only if it is restricted to this app (Google Cloud console)'), text);
166
- each(/data:image\/[a-z+]+;base64,[A-Za-z0-9+/=]{8000,}/g, m => add('✗', m.index, m.index + m[0].length, `a ${Math.round(m[0].length * 0.75 / 1024)} KB base64 image in code: put it in an asset file`), text);
172
+ // measured without a regular expression, whose repetition overflows the stack on an image of a few MB
173
+ for (let i = text.indexOf('data:image/'); i !== -1; i = text.indexOf('data:image/', i + 1)) {
174
+ const head = /^data:image\/[a-z+]+;base64,/.exec(text.slice(i, i + 40));
175
+ if (!head) continue;
176
+ let j = i + head[0].length;
177
+ while (j < text.length && BASE64.has(text.charCodeAt(j))) j++;
178
+ if (j - i >= 8000) add('✗', i, j, `a ${Math.round((j - i) * 0.75 / 1024)} KB base64 image in code: put it in an asset file`);
179
+ i = j;
180
+ }
167
181
  if (expoSdk >= 56 && deps['expo-router']) {
168
182
  each(/(['"])@react-navigation\/[^'"\n]+\1/g, m => add('✗', m.index, null, `${m[0]} doesn't work alongside Expo Router from SDK 56: import from expo-router (see the expo-router skill)`));
169
183
  }
@@ -246,6 +260,81 @@ function lineAt(code, i) {
246
260
  return n;
247
261
  }
248
262
 
263
+ // ---------- Maestro flows ----------
264
+ // a flow without its comments (each comment character a space, so positions stay): # outside quotes starts one
265
+ function stripYamlComments(text) {
266
+ return text.split('\n').map(l => {
267
+ let quote = null;
268
+ for (let i = 0; i < l.length; i++) {
269
+ const c = l[i];
270
+ if (quote) { if (c === quote) quote = null; } else if (c === '"' || c === "'") quote = c;
271
+ else if (c === '#' && (i === 0 || /\s/.test(l[i - 1]))) return l.slice(0, i) + ' '.repeat(l.length - i);
272
+ }
273
+ return l;
274
+ }).join('\n');
275
+ }
276
+ const TAPS = /^(tapOn|doubleTapOn|longPressOn)$/;
277
+ const SELECTS = /^(tapOn|doubleTapOn|longPressOn|assertVisible|assertNotVisible|scrollUntilVisible|extendedWaitUntil|copyTextFrom)$/;
278
+ const ASSERTS = /^(assertVisible|assertNotVisible|assertTrue|assertWithAI|assertNoDefectsWithAI)$/;
279
+ // buttons of system dialogs (permission prompts, alerts), which have no testID
280
+ const SYSTEM_BUTTON = /^(allow|allow while using (the )?app|allow once|only this time|while using the app|ok|cancel|continue|don.t allow|not now|deny|close)$/i;
281
+ const unquote = v => { v = v.trim(); return /^'.*'$/.test(v) ? v.slice(1, -1).replace(/''/g, "'") : /^".*"$/.test(v) ? v.slice(1, -1) : v; };
282
+ let noMaestro = false;
283
+ function checkFlow(file, text) {
284
+ const out = [];
285
+ const add = (level, at, message, always) => out.push({ level, at, end: at, message, always });
286
+ const dir = path.dirname(file);
287
+ let offset = 0;
288
+ let cmd = null;
289
+ let optional = false;
290
+ let target = '';
291
+ let asserts = 0;
292
+ const done = () => { if (cmd === 'extendedWaitUntil' && !optional) asserts++; };
293
+ for (const l of stripYamlComments(text).split('\n')) {
294
+ const at = offset;
295
+ offset += l.length + 1;
296
+ const item = /^\s*-\s*([A-Za-z]+)\s*:?\s*(.*)$/.exec(l);
297
+ if (item) { done(); cmd = item[1]; optional = false; if (ASSERTS.test(cmd)) asserts++; }
298
+ if (/^\s*optional\s*:\s*true\b/.test(l)) optional = true;
299
+ const app = /^appId\s*:\s*(.+)$/.exec(l);
300
+ if (app && !/\$\{APP_ID\}/.test(app[1])) add('!', at, `appId ${app[1].trim()}: verify.sh sets APP_ID for the build it checks, which a flavor or scheme changes: use appId: \${APP_ID}`);
301
+ // what this line selects: the command's own value (tapOn: Save), keys of an inline map, or a key below the command
302
+ const sel = [];
303
+ if (item && SELECTS.test(cmd) && item[2].trim() && !/^[{[|>]/.test(item[2].trim())) sel.push(['text', unquote(item[2])]);
304
+ for (const m of (item && /^\{/.test(item[2].trim()) ? item[2] : !item ? l : '').matchAll(/\b(text|id|visible|notVisible)\s*:\s*("[^"]*"|'(?:[^']|'')*'|[^,}\n]+)/g)) {
305
+ if (SELECTS.test(cmd || '') && unquote(m[2])) sel.push([m[1] === 'id' ? 'id' : 'text', unquote(m[2])]);
306
+ }
307
+ for (const [kind, value] of sel) {
308
+ if (TAPS.test(cmd)) target = value;
309
+ if (kind === 'text' && TAPS.test(cmd) && !SYSTEM_BUTTON.test(value) && !value.includes('${')) {
310
+ add('!', at, `${cmd} "${value}" taps by text: on iOS a label and its control both match, and on a real iPhone a hidden one can; tap by testID (id:)`);
311
+ }
312
+ if (kind === 'text' && /['’]/.test(value)) add('!', at, `"${value}" has an apostrophe, which a real iPhone can't search for (maestro-runner): match a testID, or text without it`);
313
+ }
314
+ if (item && cmd === 'inputText' && item[2].trim() && !item[2].includes('${') && /pass(word|code)?\b|\bpin\b|otp|secret/i.test(target)) {
315
+ add('!', at, `a password typed into the flow: use \${MAESTRO_PASSWORD} (or another MAESTRO_ variable) from .maestro/.env.local, which isn't committed`);
316
+ }
317
+ // a subflow or script it runs: there, or the flow fails when it gets to it
318
+ const runs = (item && /^(runFlow|runScript)$/.test(cmd) && !/^[{[|>]/.test(item[2].trim()) && item[2].trim()) || (!item && /^(runFlow|runScript)$/.test(cmd || '') && (/^\s*file\s*:\s*(.+)$/.exec(l) || [])[1]);
319
+ if (runs && !runs.includes('${') && /\.(ya?ml|js)$/.test(unquote(runs)) && !fs.existsSync(path.join(root, dir, unquote(runs)))) {
320
+ add('✗', at, `${cmd} ${unquote(runs)}: no such file next to this flow (${dir}/)`, true);
321
+ }
322
+ }
323
+ done();
324
+ for (const [re, what] of SECRETS) for (const m of text.matchAll(re)) add('✗', m.index, `${what} in a flow: anyone with the repository can read it. Revoke it, and use a MAESTRO_ variable from .maestro/.env.local`);
325
+ // a flow verify.sh runs (not a subflow) that never checks anything passes whatever the app shows
326
+ if (/^\.maestro\/[^/]+$/.test(file) && !asserts) add('!', 0, 'this flow checks nothing: add an assertVisible for what each scenario\'s THEN expects', true);
327
+ // Maestro's own syntax check (seconds, no device): AMK_MAESTRO names another maestro
328
+ const r = noMaestro ? null : spawnSync(process.env.AMK_MAESTRO || 'maestro', ['check-syntax', file], { cwd: root, encoding: 'utf8' });
329
+ if (r && r.error) noMaestro = true;
330
+ else if (r && r.status !== 0) {
331
+ const said = `${r.stdout || ''}\n${r.stderr || ''}`.split('\n').map(s => s.trim()).find(s => s && s !== 'OK') || 'invalid flow';
332
+ const n = Number((/:(\d+):\d+/.exec(said) || [])[1]) || 1;
333
+ add('✗', text.split('\n').slice(0, n - 1).reduce((a, s) => a + s.length + 1, 0), `maestro check-syntax: ${said.replace(/\s+at \/syntax-checker:\d+:\d+/, '')}`, true);
334
+ }
335
+ return out;
336
+ }
337
+
249
338
  for (const [file, lines] of changed) {
250
339
  let text;
251
340
  try { text = fs.readFileSync(path.join(root, file), 'utf8'); } catch { continue; }
@@ -253,13 +342,13 @@ for (const [file, lines] of changed) {
253
342
  const starts = [0];
254
343
  for (let i = text.indexOf('\n'); i !== -1; i = text.indexOf('\n', i + 1)) starts.push(i + 1);
255
344
  const line = i => { let lo = 0, hi = starts.length - 1; while (lo < hi) { const mid = (lo + hi + 1) >> 1; if (starts[mid] <= i) lo = mid; else hi = mid - 1; } return lo + 1; };
256
- for (const f of check(file, text)) {
345
+ for (const f of flowFile(file) ? checkFlow(file, text) : check(file, text)) {
257
346
  const from = line(f.at);
258
347
  const to = line(f.end);
259
348
  const touched = lines === 'all' || [...Array(to - from + 1).keys()].some(k => lines.has(from + k));
260
349
  const at = f.line || from;
261
350
  // on the line, or in a comment on the line above
262
- const above = /^\s*(\/\/|\/\*|\{\s*\/\*)/.test(rows[at - 2] || '') ? at - 1 : 0;
351
+ const above = /^\s*(\/\/|\/\*|\{\s*\/\*|#)/.test(rows[at - 2] || '') ? at - 1 : 0;
263
352
  const okAt = [at, above].find(n => n && /amk-ok\b/.test(rows[n - 1]));
264
353
  const ok = okAt && /amk-ok(?:\s+(\d{4}-\d{2}-\d{2}))?\s*:\s*(\S.*?)\s*(\*\/\s*\}?)?\s*$/.exec(rows[okAt - 1]);
265
354
  let message = f.message;
@@ -268,14 +357,14 @@ for (const [file, lines] of changed) {
268
357
  // kept: listed when its own lines changed; over two weeks old, asked again whenever its file changes
269
358
  if (Date.parse(today) - Date.parse(since) <= 14 * DAY) { if (touched) kept.push(`${file}:${at}: ${ok[2]} (${since})`); continue; }
270
359
  message = `kept with amk-ok since ${since}, over two weeks, and its file changed again: ${f.message}. The user's reason then: ${ok[2]}`;
271
- } else if (!touched) continue;
360
+ } else if (!touched && !f.always) continue;
272
361
  if (!findings.some(x => x.where === `${file}:${at}` && x.message === message)) findings.push({ ...f, message, where: `${file}:${at}` });
273
362
  }
274
363
  }
275
364
 
276
365
  // ---------- once for the app: an error boundary somewhere ----------
277
366
  let boundary = true;
278
- if (changed.size) {
367
+ if ([...changed.keys()].some(source)) {
279
368
  boundary = false;
280
369
  const files = [...(git('ls-files') || '').split('\n'), ...changed.keys()].filter(f => f && source(f));
281
370
  for (const f of new Set(files)) {
@@ -285,6 +374,45 @@ if (changed.size) {
285
374
  }
286
375
  }
287
376
 
377
+ // ---------- each scenario of a spec the branch touched has its steps in the spec's flows ----------
378
+ // specs/<id>/requirements.md lists them ("#### Scenario: <name>"); the flows are .maestro/<id>*.yaml, with a
379
+ // "# Scenario <n>: <name>" comment above each one's steps, as specs/_templates/flow.yaml shows
380
+ const touchedPaths = [...(sha ? (git('diff', '--name-only', '--relative', sha) || '').split('\n') : []),
381
+ ...(git('ls-files', '--others', '--exclude-standard') || '').split('\n')].filter(Boolean);
382
+ const norm = s => s.toLowerCase().replace(/[^a-z0-9]+/g, ' ').trim();
383
+ const specIds = fs.existsSync(path.join(root, 'specs')) ? fs.readdirSync(path.join(root, 'specs'), { withFileTypes: true })
384
+ .filter(e => e.isDirectory() && !/^(_|current$)/.test(e.name)).map(e => e.name) : [];
385
+ const flowsDir = fs.existsSync(path.join(root, '.maestro')) ? fs.readdirSync(path.join(root, '.maestro')).filter(f => /\.ya?ml$/.test(f)) : [];
386
+ for (const id of specIds) {
387
+ const req = `specs/${id}/requirements.md`;
388
+ let reqText;
389
+ try { reqText = fs.readFileSync(path.join(root, req), 'utf8'); } catch { continue; }
390
+ const flows = flowsDir.filter(f => f.toLowerCase().startsWith(id.toLowerCase())).map(f => `.maestro/${f}`);
391
+ if (!flows.length || !touchedPaths.some(p => p.startsWith(`specs/${id}/`) || flows.includes(p))) continue;
392
+ const blocks = flows.flatMap(f => {
393
+ const text = fs.readFileSync(path.join(root, f), 'utf8');
394
+ const code = stripYamlComments(text).split('\n');
395
+ const out = [];
396
+ text.split('\n').forEach((l, i) => {
397
+ const m = /^\s*#\s*Scenario\b\s*(\d+)?\s*[:.)-]?\s*(.*)$/i.exec(l);
398
+ if (m) out.push({ n: Number(m[1]) || 0, name: norm(m[2]), at: i });
399
+ else if (out.length) out[out.length - 1].code = `${out[out.length - 1].code || ''}${code[i]}\n`;
400
+ });
401
+ return out.map(b => ({ ...b, flow: f }));
402
+ });
403
+ [...reqText.matchAll(/^#### Scenario:\s*(.*)$/gm)].forEach((m, i) => {
404
+ const name = norm(m[1]);
405
+ const where = `${req}:${reqText.slice(0, m.index).split('\n').length}`;
406
+ const b = blocks.find(x => x.n === i + 1) || blocks.find(x => name && x.name && (x.name.includes(name) || name.includes(x.name)));
407
+ const say = message => findings.push({ level: '!', message, where });
408
+ if (!b) say(`scenario ${i + 1} "${m[1].trim()}" has no block in ${flows.join(', ')}: add "# Scenario ${i + 1}: ${m[1].trim()}" above its steps, with an assertion for its THEN and a screenshot`);
409
+ else {
410
+ if (!/-\s*(assertVisible|assertNotVisible|assertTrue|assertWithAI|extendedWaitUntil)\b/.test(b.code || '')) say(`scenario ${i + 1} "${m[1].trim()}" has no assertion in ${b.flow}: check what its THEN expects`);
411
+ if (!/-\s*takeScreenshot\b/.test(b.code || '')) say(`scenario ${i + 1} "${m[1].trim()}" has no screenshot in ${b.flow}, so it has no evidence`);
412
+ }
413
+ });
414
+ }
415
+
288
416
  // ---------- report ----------
289
417
  // by file, then line
290
418
  const order = w => { const [, file, n] = /^(.*):(\d+)$/.exec(w); return [file, Number(n)]; };
@@ -292,14 +420,15 @@ findings.sort((a, b) => { const [fa, na] = order(a.where); const [fb, nb] = orde
292
420
  const problems = findings.filter(f => f.level === '✗');
293
421
  const warnings = findings.filter(f => f.level === '!');
294
422
  const plural = (n, w) => `${n} ${w}${n === 1 ? '' : 's'}`;
295
- if (!changed.size) console.log(`rn-check: no changed source files since ${since}`);
423
+ if (!changed.size && !findings.length) console.log(`rn-check: no changed code or flows since ${since}`);
296
424
  else if (!findings.length) console.log(`rn-check: nothing to report in the lines changed since ${since}`);
297
425
  else console.log(`rn-check: ${plural(problems.length, 'problem')}, ${plural(warnings.length, 'warning')} in the lines changed since ${since}`);
298
426
  for (const f of [...problems, ...warnings]) console.log(` ${f.level} ${f.where} ${f.message}`);
427
+ if (noMaestro && [...changed.keys()].some(flowFile)) console.log(' flows not syntax-checked: Maestro isn\'t installed here');
299
428
  if (!boundary) console.log(` ! the app has no error boundary: a render error shows a blank screen or closes the app. Wrap the app in one`);
300
429
  if (kept.length) console.log(` kept with amk-ok: ${[...new Set(kept)].join('; ')}`);
301
430
  if (problems.length) {
302
- console.log(` Each ✗ fails verify.sh quick. Ask the user, for each one: fix it, or keep the line with \`// amk-ok ${today}: <their reason>\` on it.`);
431
+ console.log(` Each ✗ fails verify.sh quick. Ask the user, for each one: fix it, or keep the line with \`// amk-ok ${today}: <their reason>\` on it (# in a flow).`);
303
432
  console.log(' Never add an amk-ok without the user\'s reason. One kept over two weeks ago comes back when its file changes: ask again, and to keep it, write today\'s date.');
304
433
  }
305
434
  if (warnings.length || !boundary) console.log(' Fix each ! in the code this task changed, or tell the user why it stays.');
@@ -13,6 +13,7 @@
13
13
  # one per device, in rounds when they don't all fit, the flows split into parts when there's room
14
14
  # --spec <id> run .maestro/<id>*.yaml and save evidence in .amk/evidence/<id>/
15
15
  # --flow <file> run only this flow: a path in the project, or anywhere (an absolute path)
16
+ # --tags <list> only the flows with one of these tags in their header (tags: [smoke]), comma-separated
16
17
  # --label <n> evidence folder name (default: the spec id, or no-spec)
17
18
  # --flavor <name> Android product flavor to check (default: the first one in build.gradle)
18
19
  # --scheme <name> iOS scheme to check (default: the one named like the workspace)
@@ -33,10 +34,13 @@
33
34
  # AMK_IOS_TABLET (iPad simulator, or a connected iPad: name or UDID), AMK_ANDROID_TABLET_AVD (tablet emulator),
34
35
  # AMK_ANDROID_TABLET (serial), AMK_IOS_TEAM (the Apple team ID that signs builds for real iPhones and iPads;
35
36
  # default: the project's), AMK_PARALLEL (1 or 0: overrides plan's choice), AMK_MEMORY_GB (the memory plan
36
- # counts on: this machine's; less when other emulators stay open), AMK_DEVICES (how many devices plan checks at
37
- # once: one per 6 GB beyond 4), AMK_DEVICE_WAIT (minutes to wait for a
37
+ # counts on: this machine's; less when other emulators stay open), AMK_DISK_GB (the free disk it counts on: this
38
+ # folder's), AMK_A11Y (0: no view-tree steps on Android, so no accessibility warnings. The steps go in a copy
39
+ # of each flow, .maestro/.amk-<screen>-<flow>.yaml, where ${MAESTRO_FILENAME} is the copy's name),
40
+ # AMK_DEVICES (how many devices plan checks at once: one per 6 GB beyond 4), AMK_DEVICE_WAIT (minutes to wait for a
38
41
  # device in use, 30), AMK_LOCK_DIR (where runs on this machine claim devices; /tmp/agentic-mobile-kit-<user id>),
39
- # AMK_EVIDENCE_KEEP (how many evidence folders stay, newest first: 5; 0 keeps them all)
42
+ # AMK_EVIDENCE_KEEP (how many evidence folders stay, newest first: 5; 0 keeps them all; a prototype's and those of a
43
+ # spec in progress always stay)
40
44
  # A device in use by another run (any project on this machine) is skipped for a free one of the same kind, or waited
41
45
  # for when it was named or none is free.
42
46
  # Real iPhones and iPads (iOS 17 or later): Maestro doesn't run on them, so their flows run through maestro-runner, with
@@ -49,6 +53,7 @@ mode=${1:-all}
49
53
  args=$* # for the commands plan prints
50
54
  spec=""
51
55
  flow=""
56
+ tags=""
52
57
  label=""
53
58
  flavor=${AMK_ANDROID_FLAVOR:-}
54
59
  only=${AMK_SCREENS:-}
@@ -59,6 +64,7 @@ while [ $# -gt 0 ]; do
59
64
  case $1 in
60
65
  --spec) spec=$2; shift 2 ;;
61
66
  --flow) flow=$2; shift 2 ;;
67
+ --tags) tags=$2; shift 2 ;;
62
68
  --label) label=$2; shift 2 ;;
63
69
  --flavor) flavor=$2; shift 2 ;;
64
70
  --scheme) AMK_IOS_SCHEME=$2; export AMK_IOS_SCHEME; shift 2 ;;
@@ -86,16 +92,26 @@ case $flow in "" | /* | [A-Za-z]:/*) ;; *) flow=$root/$flow ;; esac
86
92
  evidence="$root/.amk/evidence/${label:-${spec:-no-spec}}"
87
93
  mkdir -p "$evidence"
88
94
  # Evidence stays on this machine (.amk/ is out of git): the newest AMK_EVIDENCE_KEEP folders stay, this one among
89
- # them, and older ones go (5; 0 keeps them all). A prototype's folder (<id>-prototype, material for a demo) always
90
- # stays and isn't counted
95
+ # them, and older ones go (5; 0 keeps them all). A prototype's folder (<id>-prototype, material for a demo) and those
96
+ # of a spec in progress (its progress.md isn't done: its evidence table points there) always stay and aren't counted.
97
+ # Kits before 0.7.12 gave each run without a spec a folder named after its time: those go first
91
98
  touch "$evidence"
92
99
  keep=${AMK_EVIDENCE_KEEP:-5}
93
100
  if [ "$keep" -gt 0 ] 2> /dev/null; then
94
- old=$(ls -dt "$root"/.amk/evidence/*/ 2> /dev/null | grep -v -- '-prototype/$' | tail -n +$((keep + 1)))
101
+ open=$(for p in "$root"/specs/*/progress.md; do
102
+ case $p in (*/specs/_templates/* | */specs/current/*) continue ;; esac
103
+ if [ -f "$p" ] && ! grep -qi -E '^Status:[[:space:]]*done' "$p"; then basename "$(dirname "$p")"; fi
104
+ done)
105
+ all=$(ls -dt "$root"/.amk/evidence/*/ 2> /dev/null | grep -v -- '-prototype/$' | while IFS= read -r d; do
106
+ for s in $open; do case $(basename "$d") in ("$s" | "$s"-*) continue 2 ;; esac; done
107
+ echo "$d"
108
+ done)
109
+ timed='/[0-9]{8}-[0-9]{6}/$'
110
+ old=$({ echo "$all" | grep -v -E "$timed" || true; echo "$all" | grep -E "$timed" || true; } | grep . | tail -n +$((keep + 1)))
95
111
  if [ -n "$old" ]; then
96
112
  # another run on this machine (a part, a round) may have just removed the same folder
97
113
  echo "$old" | while IFS= read -r d; do rm -r "${d:?}" 2> /dev/null || true; done
98
- echo "evidence: removed $(echo "$old" | wc -l | tr -d ' ') older folders from .amk/evidence, keeping the newest $keep (AMK_EVIDENCE_KEEP sets how many; 0 keeps them all)"
114
+ echo "evidence: removed $(echo "$old" | awk -F/ 'NR <= 5 { s = s (NR > 1 ? ", " : "") $(NF - 1) } END { print s (NR > 5 ? " and " NR - 5 " more" : "") }') from .amk/evidence, keeping the newest $keep and those of specs in progress (AMK_EVIDENCE_KEEP sets how many; 0 keeps them all)"
99
115
  fi
100
116
  fi
101
117
  status=0
@@ -103,10 +119,27 @@ status=0
103
119
  flow_files() {
104
120
  {
105
121
  if [ -n "$flow" ]; then echo "$flow"
106
- elif [ -n "$spec" ]; then find "$root/.maestro" -maxdepth 1 -iname "$spec*.yaml" 2>/dev/null | sort
107
- else find "$root/.maestro" -maxdepth 1 -name '*.yaml' 2>/dev/null | sort
122
+ elif [ -n "$spec" ]; then find "$root/.maestro" -maxdepth 1 -iname "$spec*.yaml" ! -name '.amk-*' 2>/dev/null | sort
123
+ else find "$root/.maestro" -maxdepth 1 -name '*.yaml' ! -name '.amk-*' 2>/dev/null | sort
108
124
  fi
109
- } | awk -v k="$part_k" -v n="$part_n" '(NR - 1) % n == k - 1'
125
+ } | with_tags | awk -v k="$part_k" -v n="$part_n" '(NR - 1) % n == k - 1'
126
+ }
127
+ # the flows (a path a line) with one of --tags in their header: tags: [smoke, login], or one tag a line below tags:
128
+ with_tags() {
129
+ if [ -z "$tags" ]; then cat; return; fi
130
+ TAGS=$tags node -e '
131
+ const fs = require("fs");
132
+ const want = process.env.TAGS.split(",").map(s => s.trim()).filter(Boolean);
133
+ for (const f of fs.readFileSync(0, "utf8").split("\n").filter(Boolean)) {
134
+ const head = fs.readFileSync(f, "utf8").split(/^---\s*$/m)[0].split("\n");
135
+ const at = head.findIndex(l => /^tags\s*:/.test(l));
136
+ if (at < 0) continue;
137
+ const inline = head[at].replace(/^tags\s*:/, "").trim();
138
+ const below = [];
139
+ for (let i = at + 1; i < head.length && /^\s+-\s*\S/.test(head[i]); i++) below.push(head[i].replace(/^\s+-\s*/, ""));
140
+ const tags = (inline ? inline.replace(/^\[|\]$/g, "").split(",") : below).map(s => s.trim().replace(/^["\x27]|["\x27]$/g, ""));
141
+ if (tags.some(s => want.includes(s))) console.log(f);
142
+ }'
110
143
  }
111
144
 
112
145
  # the screens to check on a platform: the ones the app supports, narrowed by --screens
@@ -154,7 +187,8 @@ claim() {
154
187
  unclaim() { l=$(lock_path "$1"); rm -f "$l/owner"; rmdir "$l" 2> /dev/null || true; held=$(echo "$held" | sed "s| $l||"); }
155
188
  release() { for l in $held; do rm -f "$l/owner"; rmdir "$l" 2> /dev/null || true; done; held=""; }
156
189
  temps="" # the temporary devices this run made for a part of the flows (--part), deleted at its end
157
- trap 'release; drop_temps' EXIT
190
+ tree_flow="" # the copy of a flow with view-tree steps (Android), next to it while it runs
191
+ trap 'release; drop_temps; [ -z "$tree_flow" ] || rm -f "$tree_flow"' EXIT
158
192
  trap 'exit 130' INT TERM
159
193
  holder() {
160
194
  l=$(lock_path "$1")
@@ -179,8 +213,17 @@ run_flow() { # screen device app_id file name orientation
179
213
  [ ! -d "$debug" ] || rm -r "$debug"
180
214
  mkdir -p "$debug" "$out/screenshots"
181
215
  target=$4
216
+ # Android: after each screenshot, Maestro saves the screen's view tree (an optional step that can't pass, half a
217
+ # second), from which a11y.js lists tappable elements without a label or under 48 dp. Maestro runs a copy of the flow
218
+ # with those steps, next to it so its subflows' paths still work, deleted after. AMK_A11Y=0 turns it off
219
+ tree_flow=""
220
+ if [ "${1%%-*}" = android ] && [ "${AMK_A11Y:-1}" != 0 ]; then
221
+ tree_flow=$(dirname "$4")/.amk-$1-$5.yaml
222
+ with_trees "$4" > "$tree_flow"
223
+ target=$tree_flow
224
+ fi
182
225
  # Maestro reads the path in the flows below itself, so it's a native path (a Windows path in Git Bash)
183
- native=$(cd "$(dirname "$4")" && { pwd -W 2> /dev/null || pwd; })/$(basename "$4")
226
+ native=$(cd "$(dirname "$target")" && { pwd -W 2> /dev/null || pwd; })/$(basename "$target")
184
227
  if is_real "$2"; then
185
228
  # on a real device every flow runs inside one, first: the first launchApp maestro-runner finds in a flow decides
186
229
  # whether it answers alerts itself, and for most it does (for permission prompts, which it can't allow ahead),
@@ -223,7 +266,8 @@ run_flow() { # screen device app_id file name orientation
223
266
  pngs "$@" > "$debug/.pngs"
224
267
  # with the app's build: on a real device, clearState installs the app again
225
268
  MAESTRO_RUNNER_NO_TELEMETRY=1 DO_NOT_TRACK=1 npx -y "$runner" --platform ios --driver wda --device "$2" \
226
- --team-id "$team" --wda-bundle-id "$3.wda" --app-file "$device_app" test -e APP_ID="$3" --output "$debug" \
269
+ --team-id "$team" --wda-bundle-id "$3.wda" --app-file "$device_app" test -e APP_ID="$3" \
270
+ -e ORIENTATION="$([ "$6" = PORTRAIT ] && echo PORTRAIT || echo landscape)" --output "$debug" \
227
271
  --flatten "$target" > "$out/$5.log" 2>&1 || rc=$?
228
272
  pngs "$@" | comm -13 "$debug/.pngs" - | while IFS= read -r f; do mv "$f" "$debug/"; done
229
273
  # its takeScreenshot files, not the ones of a step that failed (<step>-after.png)
@@ -232,9 +276,18 @@ run_flow() { # screen device app_id file name orientation
232
276
  fi
233
277
  # screenshots use relative names: older Maestro saves them in the current folder, then in --test-output-dir, and
234
278
  # from 2.7 in the debug output under <flow>/takeScreenshot/, from where they join the screen's evidence
279
+ # ORIENTATION: how this screen turned the device, for a flow to turn it back after another app's portrait-only screen
280
+ # (a call, the camera) turned it
235
281
  (cd "$out" && maestro --device "$2" test --test-output-dir "$out" --debug-output "$debug" --flatten-debug-output \
236
- -e APP_ID="$3" "$target" > "$out/$5.log" 2>&1) || rc=$?
282
+ -e APP_ID="$3" -e ORIENTATION="$6" "$target" > "$out/$5.log" 2>&1) || rc=$?
237
283
  find "$debug" -path '*/takeScreenshot/*.png' -exec cp {} "$out/screenshots/" \;
284
+ if [ -n "$tree_flow" ]; then
285
+ rm -f "$tree_flow"
286
+ tree_flow=""
287
+ # Maestro named the flow's folder after the copy, a hidden name: the flow's own, as on iOS
288
+ [ ! -d "$debug/.amk-$1-$5" ] || [ -e "$debug/$5" ] || mv "$debug/.amk-$1-$5" "$debug/$5"
289
+ node scripts/amk/a11y.js "$debug" "$density" "$5" >> "$out/.a11y" 2> /dev/null || true
290
+ fi
238
291
  return $rc
239
292
  }
240
293
 
@@ -253,6 +306,39 @@ steady() {
253
306
  [ "$in_row" -ge 3 ] || sleep 10
254
307
  done
255
308
  }
309
+ # why flow $2 failed on the device $1, in a line: the step and what Maestro said (on a real device, maestro-runner's report)
310
+ why_failed() {
311
+ if is_real "$1"; then
312
+ node -e 'try { const e = ((JSON.parse(require("fs").readFileSync(process.argv[1], "utf8")).flows || [])[0] || {}).error;
313
+ if (e) console.log(`with: ${e.split("\n")[0]}`); } catch {}' "$debug/report.json"
314
+ else
315
+ # the first step that failed is the innermost; Maestro's message comes after the subflows that failed around it
316
+ # (Run <subflow>... FAILED) and the steps onFlowComplete ran
317
+ awk '/\.\.\. FAILED[[:space:]]*$/ { step = $0; sub(/\.\.\. FAILED[[:space:]]*$/, "", step); sub(/^[[:space:]]+/, "", step)
318
+ said = ""
319
+ while ((getline line) > 0) if (line ~ /[^[:space:]]/ && line !~ /\.\.\. [A-Z]+[[:space:]]*$/) { said = line; break }
320
+ sub(/^[[:space:]]+/, "", said)
321
+ printf "at \"%s\"%s\n", step, (said == "" ? "" : ": " said); exit }' "$out/$2.log"
322
+ fi
323
+ }
324
+ # the screen when a flow failed on the device $1: the last screenshot of a failing step (Maestro's, or maestro-runner's),
325
+ # not one of the view-tree steps', which Maestro takes too
326
+ failure_shot() {
327
+ if is_real "$1"; then find "$debug" -name '*-after.png' | sort | tail -n 1
328
+ else find "$debug" -path '*/screenshots/*.png' ! -name '*amk-view-tree*' | sort | tail -n 1; fi
329
+ }
330
+ # flow $1 with a step after each screenshot (takeScreenshot: <name>) that makes Maestro save the screen's view tree,
331
+ # named after the screenshot: it waits 1 ms for an element that isn't there, and doesn't fail the flow
332
+ with_trees() {
333
+ awk '{ print }
334
+ /^[[:space:]]*-[[:space:]]*takeScreenshot:[[:space:]]*[^[:space:]#]/ {
335
+ match($0, /^[[:space:]]*/); ind = substr($0, 1, RLENGTH)
336
+ name = $0; sub(/^[[:space:]]*-[[:space:]]*takeScreenshot:[[:space:]]*/, "", name); sub(/[[:space:]]+#.*$/, "", name)
337
+ sub(/[[:space:]]+$/, "", name) # and a Windows line end
338
+ gsub(/["\047]/, "", name); gsub(/[^A-Za-z0-9_-]/, "-", name)
339
+ printf "%s- extendedWaitUntil:\n%s visible:\n%s id: \"amk-view-tree-%s\"\n%s timeout: 1\n%s optional: true\n", ind, ind, ind, name, ind, ind
340
+ }' "$1"
341
+ }
256
342
  run_flows() { # screen device app_id orientation (PORTRAIT, or LANDSCAPE_LEFT to turn it) [again]
257
343
  out="$evidence/$1"
258
344
  # a part of the flows has a folder of its own in the screen's, as the parts run at the same time: part 1 clears what
@@ -272,9 +358,14 @@ run_flows() { # screen device app_id orientation (PORTRAIT, or LANDSCAPE_LEFT to
272
358
  mkdir -p "$out"
273
359
  # this run's screenshots are the ones newer than this; the status before it, for running the screen again
274
360
  : > "$out/.started"
361
+ # each flow's result, and what the view trees showed (Android), for the report
362
+ : > "$out/.flows"
363
+ : > "$out/.a11y"
364
+ [ "${1%%-*}" != android ] || density=$(device_density "$2")
275
365
  [ -n "${5:-}" ] || before=$status
276
366
  files=$(flow_files)
277
367
  if [ -z "$files" ]; then
368
+ [ -z "$tags" ] || echo " $1: no flow has the tag $tags, so this is a check that the app opens"
278
369
  # launchApp returns before the app has drawn, and a splash screen stays up until the JavaScript has loaded:
279
370
  # a fixed wait (for text that never appears), then for the screen to settle, so the screenshot shows the app
280
371
  printf 'appId: %s\n---\n- launchApp\n- extendedWaitUntil:\n visible: "agentic-mobile-kit: waiting for the app to draw"\n timeout: 5000\n optional: true\n- waitForAnimationToEnd\n- takeScreenshot: smoke-launch\n' "$3" > "$out/smoke.yaml"
@@ -285,7 +376,8 @@ run_flows() { # screen device app_id orientation (PORTRAIT, or LANDSCAPE_LEFT to
285
376
  for f in $files; do
286
377
  total=$((total + 1))
287
378
  name=$(basename "$f" .yaml)
288
- ok=0
379
+ ok=0 why="" shot=""
380
+ : > "$out/.flow-started"
289
381
  run_flow "$1" "$2" "$3" "$f" "$name" "$4" || ok=$?
290
382
  # Maestro now and then stops before a flow starts (it printed nothing), and an emulator sometimes drops its adb
291
383
  # connection for a moment: run the flow once more (after reconnecting)
@@ -322,6 +414,19 @@ run_flows() { # screen device app_id orientation (PORTRAIT, or LANDSCAPE_LEFT to
322
414
  echo " $1: flow $name failed, see ${out#"$root"/}/$name.log and Maestro's own log ${debug#"$root"/}/maestro.log"
323
415
  status=1
324
416
  fi
417
+ if [ "$ok" != 0 ]; then
418
+ why=$(why_failed "$2" "$name" | tr '\t' ' ')
419
+ # a step the device dropped out of has no FAILED line: say so, or it reads as the flow's own failure
420
+ [ -n "$why" ] || [ "${1%%-*}" != android ] || ! lost_device "$out/$name.log" ||
421
+ why="when the device went offline, on each try (an emulator under load drops out: check again with fewer devices open)"
422
+ shot=$(failure_shot "$2")
423
+ [ -z "$why" ] || echo " failed $why"
424
+ [ -z "$shot" ] || echo " the screen when it failed: ${shot#"$root"/}"
425
+ fi
426
+ # for the report: the flow, its result, why it failed, the screenshots it took, and the screen when it failed
427
+ printf '%s\t%s\t%s\t%s\t%s\n' "$name" "$([ "$ok" = 0 ] && echo passed || echo failed)" "$why" \
428
+ "$(find "$out" -name '*.png' -newer "$out/.flow-started" | sed "s|^$evidence/||" | sort | tr '\n' ',' | sed 's/,$//')" \
429
+ "${shot#"$evidence"/}" >> "$out/.flows"
325
430
  done
326
431
  # iOS saves a turned screen's screenshots at the device's upright size, with a tag that tells viewers to turn them,
327
432
  # and some viewers ignore it: save them turned (through BMP, which keeps no tag), so every viewer shows the screen
@@ -351,6 +456,8 @@ run_flows() { # screen device app_id orientation (PORTRAIT, or LANDSCAPE_LEFT to
351
456
  fi
352
457
  shots=$(find "$out" -name '*.png' | wc -l | tr -d ' ')
353
458
  echo "$1: flows $passed/$total passed, $shots screenshots in ${out#"$root"/}"
459
+ a11y=$(grep -c . "$out/.a11y" 2> /dev/null || true)
460
+ [ "${a11y:-0}" = 0 ] || echo " $1: $a11y accessibility warning$([ "$a11y" = 1 ] || echo s) on the flows' screens (controls without a label, tap areas under 48 dp): see the report"
354
461
  if [ "$shots" = 0 ]; then
355
462
  echo " $1: no screenshots were saved, so there's no evidence; treating this as a failure"
356
463
  status=1
@@ -365,6 +472,7 @@ run_flows() { # screen device app_id orientation (PORTRAIT, or LANDSCAPE_LEFT to
365
472
  case $was in upright) other=landscape ;; *) other=portrait ;; esac
366
473
  echo " $1: the app sets its orientation in its code (${set_in% }), which can keep it $was. If it's meant to stay $was, take $other out of its native settings (UISupportedInterfaceOrientations in Info.plist, screenOrientation in AndroidManifest.xml) and this screen isn't checked any more; otherwise change that code"
367
474
  fi
475
+ echo " $1: a flow that leaves the app (a call, the camera, a picker) comes back $was when the other app's screen turned the device: turn it back after that step with - setOrientation: \${ORIENTATION}"
368
476
  fi
369
477
  }
370
478
 
@@ -644,6 +752,10 @@ device_width() {
644
752
  /Physical size/ { split($2, s, "x"); w = s[1] + 0 < s[2] + 0 ? s[1] : s[2] } /Physical density/ { d = $2 }
645
753
  END { if (d) print int(w * 160 / d) }'
646
754
  }
755
+ # the screen density (dpi) of the Android device $1: the one set for it, else its own
756
+ device_density() {
757
+ "$adb" -s "$1" shell 'wm size; wm density' 2> /dev/null | tr -d '\r' | awk -F': ' '/Override density/ { o = $2 } /Physical density/ { p = $2 } END { print o ? o : p }'
758
+ }
647
759
  avd_home=${ANDROID_AVD_HOME:-${ANDROID_USER_HOME:-$HOME/.android}/avd}
648
760
  # an AVD's folder: its .ini names it (an AVD can be renamed, and keep its folder), else <name>.avd
649
761
  avd_dir() { d=$(sed -n 's/^path=//p' "$avd_home/$1.ini" 2> /dev/null | head -n 1); echo "${d:-$avd_home/$1.avd}"; }
@@ -920,6 +1032,7 @@ plan() {
920
1032
  console.log(Math.round(os.totalmem() / 2 ** 30), Math.round(s.bavail * s.bsize / 2 ** 30))')
921
1033
  mem=$1
922
1034
  case ${AMK_MEMORY_GB:-x} in *[!0-9]*) ;; *) mem=$AMK_MEMORY_GB ;; esac
1035
+ case ${AMK_DISK_GB:-x} in *[!0-9]*) ;; *) set -- "$1" "$AMK_DISK_GB" ;; esac
923
1036
  have="$mem GB of memory, $2 GB free disk, Maestro $(maestro_version)"
924
1037
  # the devices to check, phones first and the platforms mixed, so a round of two has one of each
925
1038
  # (each platform's screens are read first: grep -q stops reading at its first match, and a writer still going then
@@ -1019,5 +1132,11 @@ case $mode in
1019
1132
  *) echo "verify: unknown mode $mode (quick, ios, android, all, build or plan)"; exit 2 ;;
1020
1133
  esac
1021
1134
 
1135
+ # the run's report, from every screen checked so far: other devices, parts and rounds add theirs
1136
+ case $mode in
1137
+ ios | android | all) if [ -n "$(find "$evidence" -name .flows 2> /dev/null | head -n 1)" ] && node scripts/amk/report.js "$evidence" > /dev/null 2>&1; then
1138
+ echo "Report: ${evidence#"$root"/}/report.md (each flow's result, why one failed, and its screenshots)"
1139
+ fi ;;
1140
+ esac
1022
1141
  echo "Evidence: ${evidence#"$root"/}/ (open the screenshots: a tool saying 'passed' isn't proof)"
1023
1142
  exit $status
@@ -1,5 +1,21 @@
1
1
  # <ID>: design
2
2
 
3
+ ## UI/UX level
4
+
5
+ <!-- For a feature with screens: the level the user confirmed before the design, and the date (n/a without screens).
6
+ Simple (the default): clean, minimal and functional. The app's own components and theme, or the platform's
7
+ standard ones; a clear hierarchy with room around it; standard navigation, lists and forms; motion only where
8
+ the platform gives it (screen transitions, pressed states).
9
+ Mixed: Simple, with modern touches where they help: cards with rounded corners and soft shadows, an accent colour
10
+ and icons, subtle transitions and feedback when something changes, skeletons while loading, friendly empty and
11
+ error states.
12
+ Premium: polished and visually engaging, following current iOS and Android design (Apple's Human Interface
13
+ Guidelines, Material 3): one type scale, spacing and colour tokens, dark mode, purposeful motion (shared
14
+ transitions, gestures), haptics on key actions, illustrations for empty states. It takes longer, and may need a
15
+ library (animation, haptics): ask before adding one.
16
+ Every level meets the same requirements and the screen rules (accessibility, large text, Reduce Motion, both
17
+ platforms): the level changes how the screens look and move, not what they do. -->
18
+
3
19
  ## Existing code to reuse
4
20
 
5
21
  <!-- Search first: components, hooks, API clients and helpers this change will use instead of writing new ones. -->
@@ -1,11 +1,19 @@
1
1
  # Copy to .maestro/<id>.yaml. verify.sh runs it on iOS and Android and keeps the screenshots as evidence.
2
2
  # Tap by testID (give every interactive element one); tapping by text can hit a label instead of the control.
3
+ # Each scenario of specs/<id>/requirements.md gets a "# Scenario <n>: <name>" comment above its steps, an assertion for
4
+ # its THEN and a screenshot: rn-check (in verify.sh quick) checks that, and Maestro's syntax, before any build.
5
+ # Steps several flows share (logging in, skipping onboarding, a permission prompt) and resetting test data go in
6
+ # .maestro/subflows/ (specs/_templates/subflow.yaml), run with "- runFlow: subflows/<name>.yaml".
3
7
  # Test logins: ${MAESTRO_EMAIL}, ${MAESTRO_PASSWORD} and the like, set in .maestro/.env.local, which isn't committed.
4
8
  # On a real iPhone (maestro-runner): text with an apostrophe can't be searched for (use a testID), and permission
5
9
  # prompts aren't answered for the flow (tapOn: { text: "Allow", optional: true }). setAirplaneMode works only on an
6
10
  # iPhone in English (it looks for "Airplane Mode" in Settings) and leaves Settings in front: launchApp: { stopApp: false }
7
11
  # brings the app back. verify.sh opens Settings on its main page before such a flow, and turns airplane mode off after.
12
+ # A flow that leaves the app (a call, the camera, a system picker) can come back upright on a turned screen: the other
13
+ # app's portrait-only screen turned the device. Turn it back after that step with "- setOrientation: ${ORIENTATION}",
14
+ # which verify.sh sets to the screen's (run by hand: maestro test -e ORIENTATION=PORTRAIT <flow>).
8
15
  appId: ${APP_ID} # set by verify.sh
16
+ # tags: [smoke] # optional: verify.sh --tags smoke checks only the flows with one of the tags
9
17
  ---
10
18
  - launchApp:
11
19
  clearState: true
@@ -0,0 +1,36 @@
1
+ # Copy to .maestro/subflows/<name>.yaml for steps several flows share: logging in, skipping onboarding, answering a
2
+ # permission prompt. A flow runs it with "- runFlow: subflows/<name>.yaml". verify.sh runs only the flows directly in
3
+ # .maestro/, not the subflows below it. This one logs in with the test account from .maestro/.env.local (MAESTRO_EMAIL,
4
+ # MAESTRO_PASSWORD), which isn't committed.
5
+ appId: ${APP_ID}
6
+ ---
7
+ - tapOn:
8
+ id: "login-email"
9
+ - inputText: ${MAESTRO_EMAIL}
10
+ - tapOn:
11
+ id: "login-password"
12
+ - inputText: ${MAESTRO_PASSWORD}
13
+ - tapOn:
14
+ id: "login-submit"
15
+ - extendedWaitUntil:
16
+ visible:
17
+ id: "home-screen"
18
+ timeout: 15000
19
+
20
+ # Test data: so a flow starts from the same data every run, reset it through the app's API first. Put a script like
21
+ # this in .maestro/subflows/reset-data.js, and run it at the start of the flow:
22
+ # - runScript:
23
+ # file: subflows/reset-data.js
24
+ # env:
25
+ # API_URL: ${MAESTRO_API_URL}
26
+ # TEST_TOKEN: ${MAESTRO_TEST_TOKEN}
27
+ # Maestro runs it with its own JavaScript, which has http (docs.maestro.dev: Make HTTP requests) and reads the env
28
+ # values as plain variables. MAESTRO_API_URL and MAESTRO_TEST_TOKEN go in .maestro/.env.local. Point it at the app's
29
+ # own test-only endpoint and test account, never at production data.
30
+ #
31
+ # /* global http, API_URL, TEST_TOKEN */
32
+ # const response = http.post(`${API_URL}/test/reset`, {
33
+ # headers: { Authorization: `Bearer ${TEST_TOKEN}`, 'Content-Type': 'application/json' },
34
+ # body: JSON.stringify({ account: 'e2e' }),
35
+ # });
36
+ # if (!response.ok) throw new Error(`resetting the test data failed: ${response.status} ${response.body}`);