docs: kane-cli as one E2E package; real devices and smart variables (0.9.0+)

[siddhant573]

Summary

Presents kane-cli as one E2E package (generate, execute, prove), and documents two features that ship in kane-cli 0.9.0 and later: real-device runs, and smart variables in objectives. The changes cover the user guide, the kane-cli agent skill (all three copies, kept identical) and the Kiro steering files.

Real devices (--target android | ios)

• New page: docs/user-guide/mobile/real-device.md.

  • Plan requirements
    • KaneAI Max: one real device.
    • KaneAI credits + Real Device Plus Automation Cloud: as many devices as the plan's parallel tests allow. Credits are account-wide.
    • Over the limit: an extra session queues for up to 10 minutes, then fails with failed to open the mobile hub session: <reason>.
  • Also covers: quick start, choosing a device (name or pattern, version by major), the app (.apk/.aab/.ipa/APP… id), the TUI live view, permission and alert settings, _test.md, testrun locally and with --remote (Linux runners, Devices: block), limitations (no adaptive heal, no no_reset) and troubleshooting.

• Updated pages: mobile/overview.md (now five targets, with a comparison table), remote-execution.md, testrun.md, evidence.md (device and device_slug), testmd/overview.md, configuration.md, running-tests.md, and the READMEs.

• Skill and Kiro: a new Real devices section in references/mobile.md with three agent rules:

  • Mention the plan once before the first real-device run.
  • Never retry when the hub refuses a session.
  • Never run --parallel above 1 on KaneAI Max.

  It also adds the real-device error lines to parsing.md, plus small additions to the testrun, evidence, testmd and setup-and-config references.

Smart variables and objective suggestions

• variables-and-context.md gets a new Runtime namespaces section ({{global.*}}, {{environment.*}}, {{smart.*}}) and a table of the smart variables kane-cli resolves reliably: dates, random data and user_name.
• The objectives cookbook (§6) and the Kiro run steering get "fresh test data on every run" patterns, plus guidance on when to suggest a smart variable instead of a fixed value.
• running-tests.md and SKILL.md examples now use {{smart.random_email}} and {{smart.next_day}}. This also corrects the old wording that {{name}} is only for global variables and secrets.

Positioning: one E2E package (generate, execute, prove)

• README: a new tagline and a three-pillar section at the top.

  • Generate: requirements → use-cases → acceptance criteria → scenarios → tests.
  • Execute: a targets table covering desktop web, the Android emulator, the iOS simulator, and real Android and iOS devices.
  • Prove: evidence packs, cover and maintain reconcile.

  "Three ways people use it" becomes "Where it fits", and the old requirements section is merged into the pillars.

• Skill (all three copies): the description (923 of 1,024 characters) and the intro now open with the three pillars.

• User-guide index: regrouped under Generate / Execute / Prove, with the targets table. getting-started.md points to the mobile and test-generation guides.

• Kiro POWER.md, skill-installer/package.json and context7.json: matching name, description and keywords.

Not included

• CHANGELOG.md: the release process writes it.
• Smart variables that don't give reliable values in kane-cli runs: the location placeholders, and the os, browser and mobile-only variables.

Checks

• diff -r shows the three skill copies (.claude/, .agents/, skill-installer/skills/) are identical.
• skill-installer npm test: 73 pass, 0 fail, 4 skipped.
• All relative links and anchors in the changed docs resolve.

🤖 Generated with Claude Code