codex-pro-agent-skills

Use when designing, reviewing, validating, debugging, or improving LLM evals and agent workflow evals. Trigger for eval-driven development, task-specific eval datasets, graders, rubrics, LLM-as-judge, human calibration, traces, transcripts, tool-call evaluation, argument validation, guardrail evaluation, handoff evaluation, multi-turn agent tasks, capability evals, regression evals, pass@k/pass^k, model/prompt/tool upgrades, continuous evaluation, and production feedback loops for LLM or agent systems.

Use when framing, reviewing, validating, or communicating product engineering decisions through business/product consulting judgment. Trigger for business value analysis, product strategy, product/market analysis, PRD/RFC/memo/deck/executive-summary writing or review, SCR/SCQA/Pyramid Principle communication, MECE/issue-tree problem structuring, UI/UX review with product/business lens, developer productivity measurement with SPACE, AI-assisted development rollout with DORA 2025, Team Topologies/org-flow/ownership reviews, risk/second-order-effect analysis, and future-facing product/system tradeoff decisions.

Use when designing, reviewing, validating, debugging, implementing, or releasing Flutter applications, especially production Flutter apps targeting iOS. Trigger for Flutter app architecture, MVVM/ViewModel/repository/service layers, Dart async and isolates, Flutter state management, widgets/navigation, platform plugins and channels, Swift Package Manager/CocoaPods iOS dependencies, Flutter persistence/cache/offline design, permissions/privacy, performance/jank/memory, testing/debugging/observability, TestFlight/App Store signing and release readiness, and long-term Flutter app maintenance.

Use when designing, reviewing, validating, debugging, implementing, or releasing production iOS applications built with Flutter, SwiftUI, UIKit, or mixed Flutter/native code. Trigger for iOS app architecture, Flutter MVVM/ViewModel/repository/service design, Swift/native architecture, SwiftUI state and navigation, platform plugins and channels, Swift Package Manager/CocoaPods dependency issues, SwiftData and local persistence, async/concurrency/lifecycle, permissions/privacy/security, performance/memory/battery, testing/debugging/observability, TestFlight/App Store signing and release readiness, and long-term iOS app maintenance.

Use when designing, reviewing, validating, debugging, or improving ML-based systems and ML product architecture. Trigger for ML system design interviews, design docs, architecture reviews, problem framing, build-vs-buy decisions, metrics/loss selection, data and labeling plans, validation/leakage/split strategy, baseline planning, error analysis, training pipelines, feature stores, A/B experiments, integration/API/release plans, serving/inference optimization, monitoring, drift handling, reliability, ownership, and long-term maintenance.

Use when designing, reviewing, validating, debugging, or improving products, product discovery, UI/UX, AI/ML product experiences, design systems, product metrics, business outcomes, assumption tests, user research plans, opportunity-solution trees, feature proposals, product requirements, UX reviews, launch/readiness plans, governance, adoption, and long-term product maintenance.

Use when designing, reviewing, validating, debugging, or improving classic software QA strategy and automated test suites. Trigger for autotests, unit tests, integration tests, contract tests, end-to-end tests, test pyramid decisions, flaky tests, test doubles, mocks, fakes, stubs, test coverage, test maintainability, CI/CD quality gates, release validation, exploratory testing, and production feedback for non-LLM software systems.

Use when designing, reviewing, validating, refactoring, testing, documenting, migrating, or operating long-lived software systems through Google-style software engineering practices. Trigger for maintainability, code health, API compatibility, Hyrum's Law, tradeoff analysis, team knowledge sharing, engineering productivity metrics, style guides, code review, documentation, unit tests, test doubles, larger tests, version control, branching, build systems, dependency management, code search, static analysis, deprecation, large-scale changes, CI/CD, release safety, managed compute, and production-facing software design.

Use when designing, reviewing, validating, debugging, operating, or improving production systems through SRE and reliability engineering practices. Trigger for SLO/SLI/SLA, error budgets, observability, monitoring, logs, traces, alerting, incident management, postmortems, on-call, reliability architecture, Kubernetes reliability, chaos/load testing, toil automation, business reliability tradeoffs, production-readiness reviews, and long-term service ownership.

Use when designing, reviewing, validating, debugging, implementing, or releasing native iOS applications and modules built with Swift, SwiftUI, UIKit, SwiftData, or Apple frameworks. Trigger for Swift language architecture, value vs reference semantics, ARC and memory ownership, SwiftUI state and navigation, UIKit interoperability, SwiftData persistence, URLSession/networking, async/await, actors, MainActor, Sendable, app lifecycle/background work, permissions/privacy/security, Keychain, performance/Instruments, XCTest/XCUITest, TestFlight/App Store signing and release readiness, and long-term native iOS app maintenance.

Use when designing, building, reviewing, validating, debugging, deploying, or operating Telegram Mini Apps and Telegram Web Apps. Trigger for Telegram bot + mini app architecture, launch modes, BotFather setup, initData authentication, backend validation, HMAC/Ed25519 checks, tma.js, Telegram WebApp API, native buttons, theme/viewport/safe area, CloudStorage/DeviceStorage/SecureStorage, biometrics, sensors, location, Telegram Stars/payments, WebView testing, mobile debugging, production readiness, security reviews, and long-term maintenance.

Use when designing, running, adapting, reviewing, or debugging a multi-LLM council workflow inspired by karpathy/llm-council. Trigger for parallel first-pass model answers, anonymized peer review and ranking, chairman/synthesizer responses, OpenRouter model configuration, transparent tabbed review UIs, ranking parsers, multi-model answer quality comparisons, and local FastAPI/React council apps.