Skip to content

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[0.10.0] - 2026-03-30

🗑ī¸ Removed

  • Schema v1 Support — Completely removed legacy v1 schema, migration tooling, and all dual-version routing logic.
  • Deleted scenario_schema.json (v1), migrate.toml, MIGRATION.md, docs/guides/migrate.md
  • Removed deprecation phase checks from generate.toml and execute.toml
  • Removed v1 detection/routing from executor skill template
  • Removed /mobile-automator:migrate command

🔄 Changed

  • Schema Rename — Renamed scenario_schema_v2.json → scenario_schema.json and docs/reference/schema-v2.md → docs/reference/schema.md (the _v2 suffix was vestigial with only one schema version).
  • Result Schema — step_id and assertion_id are now string-only (previously accepted both integer and string for v1 backward compatibility). schema_version changed from enum: ["1.0", "2.0"] to const: "2.0".
  • Install Script — scripts/install-skills.js no longer copies the v1 schema to workspace.
  • CI Validation — .github/workflows/validate-schemas.js updated to reference scenario_schema.json instead of scenario_schema_v2.json.
  • Documentation — Removed all v1/v2/legacy/deprecated/migration references across 30+ files. All "v2" qualifiers dropped — the schema is now just "the schema".

✅ Kept

  • $schema_version: "2.0" field preserved in all schemas for future extensibility.
  • CHANGELOG.md historical entries unchanged — they are a record of what happened.

[0.9.0] - 2026-03-28

✨ Added

  • Auto-Detect Business-Critical Paths — Setup Section 5.0 now uses the @codebase_investigator subagent to automatically identify critical user paths from the codebase before asking the user.
  • Analyzes navigation graphs, route definitions, screen/Activity/Fragment definitions, feature modules, and README documentation
  • Presents detected paths for user confirmation (yesno prompt) instead of requiring manual input
  • Falls back to manual text input if auto-detection finds nothing or the user rejects the detected paths
  • Reduces setup friction — users confirm instead of typing

[0.8.1] - 2026-03-28

🐛 Fixed

  • Deterministic Skill Installation — Replaced the AI-mediated file copy/replace in setup Section 6.0 with a Node.js script (scripts/install-skills.js) that handles all template operations deterministically.
  • Fixes silent file corruption (truncation, missing sections, garbled schemas) during skill installation
  • Placeholder replacement now uses split().join() in Node.js instead of in-memory AI reproduction
  • Schema files copied byte-perfect with fs.copyFileSync instead of AI read/write
  • Runtime placeholders ({{capture_to}}, {{variable_name}}) in executor skill are now correctly preserved
  • Added scenario_schema_v2.json to template verification (was previously missing)
  • Verification now checks file existence, non-zero size, and setup-placeholder absence
  • Setup Section 6.0 reduced from ~100 lines to ~30 lines

[0.8.0] - 2026-03-27

🗑ī¸ Removed

  • TestRail Integration — Removed all TestRail MCP server configuration, environment variables, and integration code.
  • Removed testrail-mcp MCP server from gemini-extension.json
  • Removed TESTRAIL_API_KEY and TESTRAIL_DOMAIN environment variables
  • Removed testrail metadata field from scenario schema
  • Removed TestRail result syncing from executor
  • Updated documentation to remove TestRail references
  • Migration path: scenarios without TestRail metadata continue to work unchanged

[0.7.0] - 2026-03-27

✨ Added

  • Execute Command Enhancements — Major UX and structural improvements to the /mobile-automator:execute command.
  • Added --device="ID" flag to bypass the interactive device selection menu and target a specific device.
  • Added --all flag to unconditionally execute all available scenarios without manual selection prompts.
  • Fully integrated structured ask_user tool calls natively for all interactive prompts (device selection, execution menu, confirmations).
  • Refined deprecation logic to evaluate v1 schema warnings early, before establishing device connections or running preconditions.

[0.6.0] - 2026-03-20

✨ Added

  • Environment Persistence — The generate command now remembers your last-used environment and skips the prompt on subsequent runs.
  • First run (or after clearing preferences): interactive prompt asks for environment and saves the selection to mobile-automator/generate_preferences.json
  • Subsequent runs: saved environment is used automatically with an ℹī¸ notice
  • --set-environment="X" — use environment X and save it as the new default preference
  • --environment="X" — one-time override, uses X for this run only without changing the saved preference
  • Stale preferences (environment removed from config) are detected automatically with a ⚠ī¸ warning and the prompt re-appears
  • Single-environment projects skip the prompt entirely (always has, now also ignores flags cleanly)

[0.5.0] - 2026-03-13

✨ Added

  • Test Report Command — New /mobile-automator:report command generates aggregated test execution reports.
  • Supports multiple output formats: table (terminal), JSON, HTML
  • JUnit XML export for CI integration
  • Shows pass rate, failed scenarios, flaky steps, average duration
  • Filter by recent runs with --last N option (default: 10)

[0.3.1] - 2026-02-25

🔄 Changed

  • Native Prompts — Migrated user interaction prompts across setup and generate flows to use the structured ask_user tool natively.

[0.3.0] - 2026-02-24

✨ Added

  • Tag-Based Filtering — Organize, filter, and execute specific subsets of your scenarios using tags.
  • Added interactive prompt for tags during test scenario generation.
  • Added --tag filter to execute command with support for AND (smoke,critical), OR (smoke|regression), and NOT (!flaky).
  • Added new /mobile-automator:list-tags command to view the tag registry and counts across the testing suite.
  • Interactive execution menu now groups scenarios by primary tags when run without arguments.
  • Strict format validation enforcing lowercase alphanumeric + hyphens across schemas and commands.

[0.2.0] - 2026-02-23

✨ Added

  • Expanded Assertion Types — Supported assertion types increased to 27, organized with tiered categorization to handle simpler and complex edge cases.
  • New != Comparison Operator — Added support for the non-equality (!=) operator in assertion comparisons.
  • Two-Pass Semantic Intent Model — Introduced a two-pass workflow for parsing test generation instructions to yield more reliable test assertions.
  • Schema Validation CI Workflow — Added a GitHub Actions workflow that automatically validates the JSON syntax of all schemas, ensures Draft-07 conformance, and validates prototype scenarios against scenario_schema_v2.json.
  • TestRail Integration — Bi-directional sync with TestRail test case management
  • Fetch test cases from TestRail via natural language step format with automation hints
  • Auto-convert TestRail steps to mobile-automator scenario JSON
  • Automatically sync test execution results back to TestRail
  • Includes device info, observations, and screenshots in TestRail test runs
  • New MCP server: testrail-mcp for TestRail API access
  • Optional testrail metadata field in scenario_schema_v2.json for 1:1 case mapping
  • Environment variables TESTRAIL_API_KEY and TESTRAIL_DOMAIN for project-scoped configuration

🔄 Changed

  • Clarified Assertion Behaviors — Improved documentation and instructions around element_visible and list_item_count assertions.
  • Skill Categories — Reindexed skill categories within the project framework.
  • Documentation Updates — Updated Gemini CLI installation source instructions in CONTRIBUTING.md and refreshed the "Last updated" date in ROADMAP.md.
  • Generator Skill Enhanced — Detects and handles TestRail URLs for automated test case fetching
  • Executor Skill Enhanced — Syncs results back to TestRail when scenario has testrail metadata

✅ Backward Compatible

  • Manual test generation unchanged — TestRail is entirely optional
  • Existing scenarios without testrail metadata work exactly as before

[0.1.1] - 2026-02-18

✨ Added

Schema v2 — Smarter, More Reliable Test Scenarios

  • New default scenario format — all generated scenarios now use schema v2 with $schema_version: "2.0" as a required root field
  • Named step IDs — steps use descriptive snake_case string IDs (e.g., "id": "tap_login") instead of integers; screenshots are now named accordingly (step_tap_login.png vs step_3.png)
  • Smart wait actions — wait_for_element, wait_for_element_gone, wait_for_loading_complete replace fixed-time wait; eliminates the #1 source of test flakiness
  • Optional steps — optional: true + on_failure: "skip" handles non-deterministic UI elements (promotional dialogs, permission prompts) without failing the test
  • Conditional steps — condition field allows steps to be skipped based on device API level, runtime state, or whether a previous step was skipped
  • Retry logic — on_failure: "retry" + retry_policy: {max_attempts, backoff_ms} distinguishes real bugs from transient failures
  • Variable capture — capture_value action + variables root block captures dynamic values (prices, IDs, amounts) for cross-step verification
  • Nested conditional sub-flows — sub_steps array expresses branching flows (e.g., "add address only if none exists") without requiring separate scenario files
  • New assertion types — pattern_match (regex), value_matches_variable, element_count (with operators), visual_state, text_changed (state transitions like "Redeem" → "Redeemed")
  • Structured preconditions — preconditions object with app_state, device_actions, device_properties enables automated pre-test setup
  • Clean metadata — execution-time fields (device_model, api_level, timestamp) removed from scenario metadata; they belong in the result JSON

/mobile-automator:migrate — Interactive v1 → v2 Migration

  • New command to upgrade existing v1 scenarios with a guided, human-supervised process
  • Automatically converts: $schema_version, integer step IDs → named strings, assertion IDs, after_step_id references, metadata cleanup, preconditions restructuring
  • Interactively handles ambiguous cases (fixed-time wait actions — asks what to wait for)
  • Always creates a .v1.bak backup before writing any changes
  • Lists what must be added manually after migration (intent the tool can't infer)

MIGRATION.md — Institutional Migration Guide

  • 6 before/after JSON examples covering every major v1 → v2 change
  • Clear breakdown of what the tool handles automatically vs. what requires human review
  • Deprecation timeline table

scenario_schema_v2.json — Formal JSON Schema

  • JSON Schema Draft-07 definition for all new scenario fields
  • Validated against both prototype scenarios with zero errors

🔄 Changed

  • Generator SKILL.md — updated to always produce v2 scenarios; expanded step translation guide (14 actions), pattern detection guide for smart waits, optional steps, conditional steps, retry, data capture, and nested sub-flows
  • Executor SKILL.md — updated with full v2 execution path: condition evaluation, variable capture, sub-flow execution, retry logic, all 9 assertion types
  • Result schema — extended additively with optional v2 fields (schema_version, captured_variables, step-level retry_count, step_duration_ms, condition_evaluated); step_id accepts both integer (v1) and string (v2)
  • generate.toml — deprecation check when --schema-version 1.0 flag is used
  • execute.toml — deprecation check when a v1 scenario is detected
  • setup.toml — copies scenario_schema_v2.json to workspace alongside v1 schema
  • GEMINI.md — dual schema registry (v2 default + v1 legacy), expanded tool mapping table with all 14 action types
  • README.md — updated schema section to describe v2, added /mobile-automator:migrate to commands table, updated project structure tree with named screenshot IDs and scenario_schema_v2.json
  • CLAUDE.md — updated file structure, schema documentation, and namespace list to reflect all new files

⚠ī¸ Deprecation Notice

Schema v1 is deprecated as of 2026-02-17. The 12-month deprecation timeline:

Phase Period Behavior
Phase 1 Now → 2026-08-17 Non-blocking warning on v1 generate/execute
Phase 2 2026-08-17 → 2027-01-17 New v1 generation blocked; execution requires acknowledgment
Phase 3 2027-01-17+ Hard fail — v1 scenarios will not execute

Migrate existing scenarios with /mobile-automator:migrate <scenario_id>. See MIGRATION.md for the full guide.

[0.1.0] - 2025-02-13

🎉 First Beta Release

Mobile Automator's first beta release brings intelligent mobile QA automation to Gemini CLI.

Core Features

  • 7-Section Setup Workflow — Comprehensive project analysis with platform detection, environment discovery, package inference, and automatic skill installation
  • Natural Language Test Generation — Describe tests in plain English, generator creates structured JSON scenarios
  • Intelligent Test Execution — AI-powered test runner with flakiness detection, regression spotting, and semantic visual testing
  • 3-Tier Architecture — Extension commands, workspace skills, and mobile-mcp automation engine

Platform Support

  • Android (native, Kotlin, Java)
  • iOS (native, Swift, Objective-C)
  • Flutter
  • React Native
  • Kotlin Multiplatform (KMP)
  • Compose Multiplatform (CMP)

Test Capabilities

  • 7 action types (launch_app, tap, type, swipe, press_button, wait, open_url)
  • 4 assertion types (screenshot_match, element_exists, element_text, element_not_exists)
  • Automatic precondition handling
  • Environment-aware package selection
  • Flakiness detection with automatic retry
  • Semantic visual testing (AI vision, not pixel matching)

Back to Docs | View on GitHub