happyword

Use this skill when the user asks to release the iOS app, upload a build, submit to App Store Connect review, or "在 App Store 上架/提交审核" for WordMagicGame.

Seed a lesson-import draft on a deployed Happyword API (Preview or prod) and optionally run the extract-pending cron so it reaches `pending` (待复核). Uses tools starting with `hw_`. Invoke when the user wants to upload a textbook photo for review against a specific Vercel URL, mirror E2E `test_lesson_import_cron_e2e`, or smoke-test lesson import + cron on a branch deployment.

Investigate why a Vercel preview / production deployment of the FastAPI server is broken (404 on every path, 500 on every path, e2e CI red, etc.). Pulls the right log surface for the symptom and maps HTTP error headers to root causes. Use whenever the deployed server returns wrong responses, the server-ci e2e job fails, or autofix loops keep retrying without converging.

Health-check every active Vercel Preview deployment listed in the live manifest at `https://happyword.cool/api/v1/public/preview-urls.json`. Probes `GET /api/v1/public/health` through Vercel Deployment Protection (`x-vercel-protection-bypass` header), prints a one-line status per preview, and exits non-zero if any are sick. Use when the user asks to "test preview health", "check previews", "are PR previews alive?", before merging anything that touches `server/`, after a fleet-wide redeploy or env-var rotation, or after rotating `VERCEL_AUTOMATION_BYPASS_SECRET` — do NOT loop ad-hoc `curl` calls or use `tools/vercel/smoke-prod.sh` (production-only) for this.

Drives a per-feature visual + spec-anchored gap scout across HarmonyOS / iOS / Android using tools/parity_scout/. Use when asked to "find iOS / Android gaps vs HarmonyOS main", "check parity for <feature>", or "screenshot the three platforms and tell me what's different".

Drives a per-feature visual + spec-anchored gap scout across HarmonyOS / iOS / Android using tools/parity_scout/. Use when asked to "find iOS / Android gaps vs HarmonyOS main", "check parity for <feature>", or "screenshot the three platforms and tell me what's different".

Use when investigating iOS or Android parity gaps against the HarmonyOS baseline in WordMagicGame, especially when comparing behavior, UI style, stable IDs, screenshots, specs, plans, or three-platform feature docs.

Schedules the per-feature lifecycle for HarmonyOS / iOS / Android: intake, design, Harmony build+test, stabilization gate with human approval, parallel iOS+Android replication, parity checklist. Use when adding a new product feature, replicating an approved feature to iOS or Android, or asking "what stage is this feature in?"

Schedules a HarmonyOS autofix loop (build, codelinter, no-device unit, emulator, UI test) with guards and failure classification. Use when the user wants to fix until green, run a local CI-style loop, hvigor until tests pass, or asks for an orchestrated HarmonyOS build-and-test cycle with human escalation on infra failures.

Runs OHPM and Hvigor build steps for this HarmonyOS repo using the project manifest. Use when compiling the app, after dependency changes, or as the first step of an autofix loop.

Runs HarmonyOS CodeLinter (codelinter) on the project after a successful HAP build, using code-linter.json5, and fixes reported issues. Use in autofix loops immediately after harmony-build, or when addressing static analysis / code style findings.

Verifies or starts a HarmonyOS simulator or USB device via hdc, without running UI tests. Use after no-device unit tests in an autofix loop, or when hdc list targets is empty before on-device steps.

Given a classified failure tier and log excerpts, proposes minimal allowed code or test changes for HarmonyOS autofix—no duplicate taxonomy. Use only after test-failure-classifier and when agent_safe or agent_risky applies; not for unguided refactors.

Gathers and summarizes failure evidence from Hvigor output, test reports, and hdc hilog in the order defined by the project manifest. Use after any failed build, unit, or UI test step in a HarmonyOS autofix flow.

Installs HAP and runs on-device or Instrument UI tests (ohosTest, UiTest) per the project manifest. Use when hdc is up and the autofix loop reaches device-dependent tests, or when debugging failed ohosTest cases.

Runs no-device local unit tests for the HarmonyOS entry module using manifest commands. Use after a successful build and **codelinter** (in a full loop) and before starting the simulator in an autofix loop, or when fixing Local test failures under entry/src/test.

Constrains shell operations during HarmonyOS local CI and autofix runs—no destructive broad deletes, no unreviewed download-pipe, prefer manifest-defined commands. Use when executing hvigor, hdc, ohpm, or automation loops that run shell in the project.

Classifies build, lint, and test failures into category (compile, lint, unit, ui, env-infra) and fixability tier only—no file edit instructions. Use after harmony-log-analyzer output or when routing HarmonyOS hvigor, codelinter, hdc, or on-device test failures in an autofix loop.

Clean up old Vercel preview deployments for the happyword server project using the repo-owned branch-prune script (dry-run then apply). Use when the user asks to prune/clean/delete Vercel deployments, reduce deployment clutter, or reset preview history — do NOT use ad-hoc `vercel rm` bulk deletes unless the user explicitly requests a different behavior.

Manually POST Vercel cron endpoints (e.g. extract-pending) using VERCEL_CRON_SECRET from ~/.env. Use when Vercel scheduled cron is unavailable, after imports stuck in extracting, verifying Production/Preview before merge, or debugging async extraction — do not expose secrets in logs or chat.

Use when generating editable SVG/vector game art, icons, characters, props, UI assets, or marketing illustrations with Recraft V4 Vector from Codex.

Use when generating editable SVG/vector game art, icons, characters, props, UI assets, or marketing illustrations with Recraft V4 Vector from Cursor.

Use when working on 2D game mechanics, physics-based gameplay, or sprite-based interactions.

Use when working on 3D game mechanics, physics simulations, or 3D environment interactions.

Use when working on aI-driven gameplay, intelligent NPCs, procedural generation, player behavior prediction, and ML-enhanced game features.

Use when working on analytics implementation, KPI tracking, A/B testing, and data-driven optimization.

Use when working on creating complex animations, optimizing animation performance, or implementing animation state machines.

Use when working on asset pipeline setup, resource loading strategies, or bundle configuration in Cocos Creator projects.

Use when working on server communication, data synchronization, cloud services integration, and backend architecture in Cocos Creator projects.

Use when working on build setup, CI/CD pipelines, or platform-specific build issues.

Use when working on casual game development, monetization strategies, or hyper-casual game design.

Use when working on creating custom components, implementing component communication patterns, or refactoring existing component architecture.

Use when working on improving install rates and user acquisition metrics.

Use when working in a Cocos Creator project and general Cocos Creator conventions, project structure, TypeScript defaults, or performance expectations should apply.

Use when working on level architecture, difficulty balancing, progression mechanics, and procedural content generation.

Use when working on mobile-specific performance issues, battery optimization, or platform-specific features.

Use when working on any multiplayer functionality, networking architecture, or online features.

Use when working on performance issues, optimization strategies, or when targeting low-end devices.

Use when working on platform adaptations, native integrations, store submissions, and cross-platform compatibility implementations.

Use when working on playable ad projects, from initial concept to final delivery.