lude-kit

Use when threat modeling a system or feature, reviewing code or a design for security flaws, hardening auth / authorization / sessions / secrets, responding to a suspected vulnerability or incident, evaluating dependencies for CVEs, classifying data sensitivity, or designing security controls (CSP, CORS, rate limiting, WAF rules, audit logging, encryption-at-rest, encryption-in-transit). Triggers: security, threat model, STRIDE, OWASP, CVE, vulnerability, secret, leak, IDOR, SSRF, XSS, CSRF, SQLi, prompt injection, supply chain, auth, authz, RBAC, encryption, KMS, secrets, compliance, SOC2, GDPR, HIPAA, PCI. Produces threat models, secure-review findings, hardening plans, incident triage notes. Authorized contexts only: defensive security, pentest engagements with scope, CTF, security research.

Use when designing a system, choosing a database / framework / cloud / message bus, writing an ADR or RFC, deciding build vs buy, planning capacity or scaling, reviewing an architecture diagram or proposal, sequencing a migration, or weighing technical tradeoffs at the CTO level. Triggers: architect, system design, HLD, high level design, ADR, RFC, topology, capacity, scaling, build vs buy, migration plan, tech selection, tradeoff. Produces ADRs, RFCs, system diagrams, capacity plans, migration sequences. Not for implementation or code review, hand off to senior-backend-engineer / senior-frontend-engineer.

Use when {trigger verbs}, {artifact nouns}, or {situations}. Produces {outputs}. {Antitrigger if needed, name the better fit skill}.

Use when designing, implementing, or reviewing automotive software for ECUs, infotainment, telematics, ADAS adjacencies, EV battery management, or V2X; when classifying hazards and assigning ASIL ratings under ISO 26262; when running a cybersecurity threat analysis under ISO 21434; when planning OTA campaigns under UN R156 or type approval under UN R155; when choosing between Classic AUTOSAR and Adaptive AUTOSAR; when designing CAN, CAN-FD, LIN, FlexRay, or Automotive Ethernet topologies with SOME/IP; when locking down UDS and DoIP diagnostics; when planning HIL, vehicle in the loop, and fleet validation for a multi year program. Triggers: automotive, vehicle, car, ECU, AUTOSAR, Classic AUTOSAR, Adaptive AUTOSAR, ISO 26262, ASIL, ASIL-A, ASIL-B, ASIL-C, ASIL-D, HARA, ISO 21434, TARA, UN R155, UN R156, CAN, CAN-FD, LIN, FlexRay, Automotive Ethernet, SOME/IP, SecOC, UDS, DoIP, OBD-II, V2X, V2V, V2I, ADAS, infotainment, IVI, Android Automotive, QNX, MISRA C, MISRA C++, telematics, OTA for vehicles, recall.

Use when building, reviewing, or operating online stores, storefronts, catalogs, carts, checkouts, inventory, order management, fulfillment, returns, and promotions. Covers product / variant / SKU modeling, PIM, external identifiers (GTIN, EAN, MPN), cart and checkout flows, pricing and promotion rule engines, tax (Avalara, TaxJar, Stripe Tax), shipping rates, OMS, ATP and reservations, RMA, fraud and chargeback workflows, peak readiness (Black Friday, drops, flash sales), and platform choice (Shopify, BigCommerce, commercetools, Magento, Adobe Commerce, headless on Next.js). Triggers: ecommerce, e commerce, store, storefront, catalog, PIM, SKU, GTIN, EAN, MPN, cart, checkout, conversion, abandoned cart, shipping, fulfillment, OMS, inventory, ATP, ATS, returns, RMA, refund, promotion, discount, coupon, loyalty, gift card, BNPL, Black Friday, flash sale, fraud, chargeback, Shopify, BigCommerce, commercetools, Magento, headless commerce. Produces catalog schemas, checkout sequences, order state machines.

Use when designing, implementing, or reviewing education technology: learning management systems (LMS), MOOCs, K-12 classroom tools, higher ed admin, assessment and proctoring, tutoring, gradebooks, parent portals, and student information system (SIS) integrations. Covers interoperability (LTI 1.3, OneRoster, xAPI, SCORM, QTI), student data privacy (FERPA, COPPA, CIPA, GDPR-K), age gating and parental consent, accessibility for educational content (WCAG 2.1 AA, captions, MathML, IEP accommodations), and classroom workflow for teachers and students on Chromebooks, school iPads, and locked down browsers. Triggers: edtech, education technology, LMS, Canvas, Moodle, Blackboard, Schoology, Google Classroom, K-12, higher ed, MOOC, tutoring, assessment, proctoring, quiz, grade, gradebook, SIS, LTI 1.3, OneRoster, xAPI, SCORM, QTI, FERPA, COPPA, CIPA, GDPR-K, age gate, student data privacy, classroom, teacher, student, parent portal, IEP, accommodation. Produces data classification tables.

Use when planning a sprint or week, breaking an epic into tickets, sizing / estimating work, sequencing tasks across people, unblocking a stuck engineer, running a standup or retro, preparing a 1:1, writing a project update, or re-prioritizing in response to a fire. Triggers: sprint, planning, tickets, breakdown, estimate, story points, standup, retro, 1:1, status update, unblock, delegate, capacity, WIP. Produces ticket breakdowns, sprint plans, status updates, retro outcomes, 1:1 agendas, project trackers. Not for technical design (use staff-software-architect) or hands on implementation (use senior-backend-engineer / senior-frontend-engineer).

Use when designing, implementing, or reviewing payments, money movement, accounts, balances, ledgers, settlement, reconciliation, KYC, AML, sanctions screening, disputes, chargebacks, FX, or regulated financial product surfaces. Covers authorization vs capture, refunds, partial refunds, refunds across days, idempotency on money endpoints, double entry posting, bank file ingestion, processor integrations, PCI DSS scope reduction, and tokenization. Triggers: fintech, payments, payment processing, card processing, ACH, wire, SEPA, Faster Payments, PayPal, Stripe, Adyen, ledger, double entry, journal, account, balance, settlement, clearing, authorization, capture, refund, chargeback, dispute, KYC, AML, PCI DSS, PCI scope, SCA, 3DS, FX, currency, reconciliation, bank file, ISO 20022, NACHA, OFAC, sanctions, BSA, transaction monitoring. Produces money flow diagrams, ledger schemas, idempotency designs, reconciliation jobs, PCI scope diagrams, KYC decision logs, dispute case shapes.

Use when building public services, benefit programs, or internal government systems that must reach every member of the public. Covers accessibility (WCAG 2.1 AA, Section 508), plain language, multilingual UI, low bandwidth and low end device support, identity proofing (IAL2, AAL2), FedRAMP / FISMA / StateRAMP authority to operate, NIST 800-53 control mapping, open standards mandates, public records / FOIA discipline, and procurement constraints. Triggers: government, gov tech, public sector, federal, state, local, civic tech, USDS, 18F, GDS, FedRAMP, FISMA, StateRAMP, ATO, authority to operate, NIST 800-53, Section 508, accessibility, WCAG 2.1, eIDAS, GDPR, GDS service standard, plain language, FOIA, public records, procurement, GSA, SAM, OMB, multilingual, low literacy, eligibility, benefits, SNAP, Medicaid, unemployment, identity proofing, IAL2, AAL2. Produces accessibility conformance reports (VPAT), WCAG test plans, plain language passes, ATO boundary diagrams.

Use when building, integrating, or reviewing healthcare software: patient portals, clinician tools, EHR integrations, telehealth, digital therapeutics, claims and eligibility flows, or anything that touches PHI / ePHI. Covers HIPAA, HITECH, 42 CFR Part 2, GDPR for health data, PIPEDA, NHS Digital, 21 CFR Part 11, SaMD (software as a medical device) scope, FHIR R4, HL7 v2, CDA, DICOM, IHE profiles, ICD-10, CPT, LOINC, SNOMED CT, NPI, X12 (270, 271, 837, 835), Epic / Cerner / athenahealth integration, Mirth interface engines, patient matching, master patient index, audit trails, break glass access, BAA inventory, minimum necessary access. Produces PHI data flow diagrams, FHIR resource maps, audit log shapes, access control matrices, BAA tracking sheets, HL7 v2 interface specs. Triggers: healthcare, HIPAA, PHI, ePHI, EHR, EMR, FHIR, HL7, CDA, DICOM, ICD-10, CPT, LOINC, SNOMED, NPI, SaMD, FDA, clinical, clinician, patient portal, Epic, Cerner, telehealth, prior authorization, BAA, covered entity.

Use when designing, operating, or scaling the cloud and edge platform behind a fleet of connected devices: provisioning identity at factory or first boot, rotating X.509 certs, choosing MQTT / CoAP / LwM2M topic and resource models, ingesting telemetry at scale, designing command and control with acks, shipping OTA updates across thousands to millions of devices in staged rings with health gates and rollback, modeling device shadows / digital twins, budgeting bytes and dollars per device per month, picking connectivity (cellular, BLE, LoRaWAN, NB-IoT, LTE-M, eSIM), and managing brokers like AWS IoT, Azure IoT Hub, EMQX, HiveMQ, ThingsBoard. Triggers: IoT, fleet, device fleet, MQTT, CoAP, LwM2M, BLE, LoRaWAN, NB-IoT, LTE-M, cellular IoT, eSIM, device provisioning, device identity, certificate provisioning, AWS IoT, Azure IoT Hub, Azure IoT Edge, GCP IoT Core, ThingsBoard, EMQX, HiveMQ, OTA at scale, staged rollout, device shadow, digital twin, telemetry, command and control, ingest, time series.

Use when designing, building, reviewing, or operating last mile delivery, freight, warehouse, fleet routing, and supply chain visibility systems. Covers vehicle routing (VRP, OR Tools, OSRM), geospatial indexing (H3, S2, GeoJSON, geofences), ETA prediction and calibration, dispatch, driver and scanner apps, warehouse management (receive, putaway, slot, pick, pack, ship), carrier integrations (FedEx, UPS, USPS, DHL, regional carriers, 3PL), EDI (EDI 856 ASN, EDI 940 warehouse shipping order), freight (LTL, FTL), tracking, proof of delivery (POD), and exception handling (OS&D, damaged, refused, lost). Triggers: logistics, last mile, delivery, fleet routing, VRP, dispatch, OSRM, OR Tools, geofence, H3, S2, GeoJSON, ETA, route optimization, driver app, scanner, warehouse, WMS, slotting, picking, packing, shipping label, carrier, FedEx, UPS, USPS, DHL, 3PL, EDI, EDI 856, EDI 940, freight, LTL, FTL, OS&D, tracking, POD, hours of service, address validation. Produces route plans, carrier abstractions.

Use when designing, building, or operating video and audio streaming products: VOD libraries, live streaming, real time conferencing, sports, music. Covers encoding pipelines, ABR ladder design, packaging (HLS, DASH, CMAF), codec strategy (H.264, H.265, HEVC, AV1, VP9, Opus, AAC), DRM (Widevine, FairPlay, PlayReady, CENC, EME, MSE), player SDKs (web, iOS, Android, smart TV, Chromecast, AirPlay), CDN strategy and cost, low latency delivery (LL-HLS, CMAF chunked, WebRTC, SRT, RTMP ingest), forensic watermarking, and QoE telemetry (startup time, rebuffer ratio, video start failure, bitrate distribution). Triggers: streaming, video, audio, live stream, VOD, transcode, encode, manifest, transmuxing, packaging, ingest, CDN cost. Produces ABR ladders, packaging plans, DRM key flows, QoE metric sets, live latency plans, CDN cost models. Not for generic backend APIs, see `senior-backend-engineer`. Not for player UI chrome, see `senior-frontend-engineer`. Not for native player SDK integration on device.

Use when designing, building, evaluating, or operating AI agents and agentic systems: single agent loops, planner / executor splits, orchestrator with subagents, swarms, tool using agents, ReAct style loops, and Model Context Protocol (MCP) servers. Covers tool surface design, agent state and checkpointing, step / token / dollar / wall time budgets, termination conditions, human in the loop interrupts, multi agent topology, agent trace schemas, eval harnesses for verifiable tasks, and layered safety gates. Triggers: agent, AI agent, autonomous agent, multi agent, orchestration, planning, ReAct, tool use, function calling, MCP, Model Context Protocol, agent SDK, Claude Agent SDK, agent loop, agent state, agent termination, subagent, swarm, agent observability, agent trace, agent eval, agent guardrail, human in the loop, agent budget, agent cost. Produces tool definitions, agent loop skeletons, multi agent topologies, trace schemas, agent eval harnesses, human in the loop interrupt specs, safety gate plans.

Use when threat modeling an LLM or agent system, defending against prompt injection (direct and indirect), designing output safety pipelines, hardening tool use authorization, running an authorized red team set, classifying a system under EU AI Act / NIST AI RMF / ISO 42001, responding to an AI safety incident (jailbreak gone public, harmful output reported, system prompt leak), or evaluating training data privacy risk. Triggers: AI safety, AI security, LLM security, prompt injection, indirect prompt injection, jailbreak, output safety, content filter, moderation, model exfiltration, prompt extraction, system prompt leak, training data extraction, agent safety, tool safety, EU AI Act, NIST AI RMF, ISO 42001, OWASP LLM Top 10, red team AI, alignment, refusal, harmful content, CSAM, NCMEC. Produces AI threat models, defense in depth diagrams, red team sets, output safety pipelines, tool authorization matrices, regulatory classification docs, AI incident response plans.

Use when designing, implementing, or reviewing backend code, APIs (REST, GraphQL, gRPC), services, workers, schedulers, queues, databases, and data models. Covers endpoint design, validation, auth, pagination, idempotency, retries, rate limiting, schema design, migrations, indexing, transactions, caching, background jobs, and observability hooks. Triggers: backend, back-end, API, endpoint, route, REST, GraphQL, gRPC, schema, migration, query, index, transaction, queue, worker, job, cron, cache, rate limit, idempotent, webhook. Produces endpoints, services, schemas, migrations, background jobs, API contracts. Not for UI work, see senior-frontend-engineer. Not for top down system topology, see staff-software-architect.

Use when designing, implementing, or reviewing smart contracts and on chain protocols, plus the off chain infrastructure (indexers, RPC, wallet integration, monitoring) around them. Covers contract specs, invariants, access control, upgradability (UUPS, transparent, beacon proxies), gas budgets, oracle integration, MEV and front running exposure, fuzz and invariant testing, mainnet fork tests, audit preparation, deployment runbooks, and indexer schemas. Stacks: EVM (Solidity, Vyper) on Ethereum and L2s (Optimism, Arbitrum, Base, zkSync, Polygon), Solana (Rust, Anchor), Sui and Aptos (Move). Triggers: blockchain, smart contract, Solidity, Vyper, EVM, Ethereum, L1, L2, rollup, Optimism, Arbitrum, Base, zkSync, Polygon, Solana, Anchor, Sui, Aptos, Move, gas, MEV, front running, slippage, oracle, ERC-20, ERC-721, ERC-1155, ERC-4626, ABI, indexer, subgraph, The Graph, Foundry, Hardhat, slither, mythril, Echidna, audit, formal verification, multisig, timelock, proxy, upgrade, dApp, wallet.

Use when designing, training, evaluating, or shipping computer vision systems: image classification, object detection, segmentation, OCR, document AI, video understanding, action recognition, tracking, pose estimation, depth, multi camera, augmented reality, edge inference. Triggers: computer vision, CV, image classification, object detection, segmentation, OCR, document AI, video understanding, action recognition, tracking, YOLO, YOLOv8, SAM, SAM-2, CLIP, DINOv2, ViT, OpenCV, image pipeline, camera calibration, vision language, multi modal, augmented reality, ARKit, ARCore, depth estimation, pose estimation, multi camera, edge inference, ONNX, CoreML, TensorRT, NPU, quantization. Produces capture plans, annotation rubrics, sliced eval sets, calibration plots, augmentation policies, and export pipelines for the target runtime. Not for the broader ML system rigor (training pipelines, registry, drift), see `senior-ml-engineer` and `senior-mlops-engineer`. Not for the eval harness platform.

Use when designing, building, reviewing, or operating data pipelines, warehouses, lakes, and lakehouses; batch and streaming ELT/ETL; dbt models; orchestration (Airflow, Dagster, Prefect, Mage); transformation (Spark, Flink, SQL); ingestion from Kafka, Kinesis, Pub/Sub, CDC; and storage on Snowflake, BigQuery, Redshift, Databricks, Iceberg, Delta. Triggers: data engineering, data pipeline, batch, streaming, ETL, ELT, dbt, Airflow, Dagster, Prefect, Spark, Flink, Kafka, Kinesis, Pub/Sub, warehouse, lake, lakehouse, Iceberg, Delta, Snowflake, BigQuery, Redshift, Databricks, SCD, late arriving data, data quality, Great Expectations, data contract, lineage, OpenLineage, freshness SLO, idempotency, watermark, exactly once, partition pruning, clustering, backfill. Produces data contracts, dbt models with tests, orchestrator DAGs, backfill plans, dataset cards, lineage wiring. Antitrigger: not for warehouse table modeling decisions in isolation (see `data-modeler`).

Use when designing an experiment or A/B test, writing or reviewing an experiment proposal or analysis plan, sizing a study (power, alpha, MDE, minimum detectable effect), checking an A/A test, picking a unit of randomization, choosing between A/B, multi armed bandit, switchback, or a quasi experiment (difference in differences, regression discontinuity, instrumental variable, synthetic control), analyzing results with confidence intervals and multiple testing correction, interpreting lift, defining primary and guardrail metrics, running cohort or segmentation analysis, or writing a result memo with a decision recommendation. Triggers: data scientist, experiment, A/B test, A/A test, hypothesis, p value, confidence interval, CI, multi armed bandit, Thompson sampling, switchback, causal, causal inference, lift, statistical power, sample size, MDE, propensity, segmentation, cohort, regression discontinuity, instrumental variable, difference in differences, synthetic control, ATE, ATT, observational study.

Use when building a sample app or demo app, writing a getting started tutorial or integration guide, preparing a conference talk or workshop, running a livestream demo or office hours, scoping a hackathon, drafting an ambassador or partner relations plan, or routing community signal back to product and engineering. Covers dogfooding the product, time to first success, activation funnels, developer experience friction, and partner platform integration walkthroughs. Triggers: developer advocate, DevRel, developer relations, evangelist, community, sample app, demo app, tutorial, getting started, quickstart, integration guide, partner, ambassador, hackathon, workshop, conference talk, office hours, livestream, twitch stream, developer experience. Produces sample apps, getting started tutorials, integration guides, talk abstracts, workshop runbooks, community feedback reports. Not for authoritative API reference or canonical docs, see senior-technical-writer. Not for product roadmap or PRD ownership.

Use when designing an eval set or eval harness for an LLM app, agent, RAG pipeline, classifier, or generative output; building a gold set; configuring an LLM as judge with a rubric; calibrating a judge against human raters; designing slice metrics; wiring a regression suite into CI; running a vibe check with rigor; choosing between exact match, BLEU, ROUGE, BERTScore, faithfulness, groundedness, or retrieval recall at K; computing inter rater agreement (Cohen kappa, Krippendorff alpha); auditing judge drift; or reporting eval deltas vs a baseline. Triggers: eval, evaluation, LLM eval, AI eval, judge, LLM as judge, gold set, holdout, regression suite, slice metric, BLEU, ROUGE, BERTScore, faithfulness, groundedness, retrieval recall, agent eval, eval harness, prompt eval, rubric, calibration, inter rater agreement, Cohen kappa, MMLU, HELM, MT Bench, AlpacaEval, vibe check, custom eval. Produces eval task specs, gold set construction plans, judge configurations, harness run reports, regression gate policies.

Use when scoping, justifying, running, evaluating, or operating a fine tune of an LLM or other foundation model: supervised fine tuning (SFT), direct preference optimization (DPO), RLHF or RLAIF, instruction tuning, continued pretraining, parameter efficient adapters (LoRA, QLoRA, PEFT), knowledge distillation, dataset curation, preference pair labeling, decontamination against eval, hosted fine tuning APIs (OpenAI, Anthropic, Together, Replicate), or bring your own GPU training on `Llama-3.1-70B`, `Mistral-Nemo`, Qwen, Gemma, HuggingFace base models. Triggers: fine tune, fine tuning, SFT, DPO, RLHF, RLAIF, instruction tuning, continued pretraining, LoRA, QLoRA, PEFT, adapter, distillation, reward model, preference pairs, dataset curation, instruction dataset, base model, foundation model, HuggingFace, Llama, Mistral, Qwen, Gemma, Together, Replicate, `bitsandbytes`, catastrophic forgetting, model card. Produces fine tune justification docs, dataset cards, training configs, eval delta reports, model cards.

Use when designing, implementing, evaluating, shipping, or operating production LLM applications: chat, copilots, classification, structured extraction, summarization, drafting, and agentic flows. Covers prompt design under eval, structured output (JSON schema, regex, grammars), tool use and function calling, retrieval integration, model selection and version pinning, streaming UX, prompt caching, cost and latency budgets, observability on every call, rollout (shadow, canary, holdout), and prompt injection defense. Triggers: LLM, large language model, GPT, Claude, Gemini, Llama, Mistral, OpenAI SDK, Anthropic SDK, prompt engineering, prompt design, prompt versioning, structured output, JSON mode, function calling, tool use, few shot, chain of thought, agent, retrieval, RAG, embedding, vector, eval, LLM eval, hallucination, jailbreak, prompt injection, AI Gateway, model routing, fallback, cost per call, token budget, streaming, caching, prompt cache. Produces versioned prompt files, LLM call wrappers.

Use when designing, training, evaluating, shipping, or operating machine learning models and LLM applications in production. Covers problem framing, eval harness design, feature store contracts, training pipelines, offline evaluation against baselines, shadow deploys, A/B rollout, drift monitoring, retraining cadence, batch and online inference with latency budgets, and LLM app systems (retrieval, structured output, fine tuning, prompt eval). Triggers: ML, machine learning, model, training, serving, inference, feature store, online inference, batch inference, embedding, vector, fine tune, retraining, model drift, evaluation, eval harness, holdout, classification, regression, ranking, recommender, retrieval, RAG, LLM app, prompt evaluation, structured output, shadow model, A/A test. Produces eval harnesses, feature contracts, training run configs, model cards, shadow and canary plans, LLM app eval rubrics. Not for research and experimentation, see `senior-data-scientist`. Not for serving platform, registry.

Use when planning, building, reviewing, or shipping mobile apps across iOS and Android. Triggers: mobile, iOS, Android, native, cross platform, React Native, Expo, Flutter, Dart, Kotlin Multiplatform, KMP, Compose Multiplatform, Capacitor, Cordova, PWA, app store, App Store Connect, Google Play, Play Console, TestFlight, internal testing, push notification, APNs, FCM, deep link, universal link, App Link, offline, sync, background fetch, BGTaskScheduler, WorkManager, low end device, energy, battery, Material Design 3, Apple HIG, signing, provisioning, staged rollout, remote config, feature flag, kill switch. Produces platform decision matrices, mobile architecture sketches, push and deep link designs, offline sync strategies, and release checklists. Distinct from swift-ios-expert (iOS dialect deep dive); this skill is cross platform and decision focused across native, React Native, Flutter, and KMP. Not for visual design from scratch, see senior-ux-designer. Not for backend API design.

Use when designing, building, or operating the gateway between applications and LLM or model providers: routing requests across Claude, OpenAI, Gemini, and open weights, enforcing per route SLOs, configuring provider failover, tracking cost per call site, designing prompt and semantic caches, applying per tenant rate limits, supporting BYOK (bring your own key), enforcing zero data retention (ZDR) and regional routing, or wiring gateway observability. Triggers: model router, AI gateway, Vercel AI Gateway, OpenRouter, LiteLLM, Portkey, model fallback, provider failover, cost routing, semantic cache, prompt cache, rate limit per tenant, BYOK, bring your own key, ZDR, zero retention, prompt logging, multi provider, provider abstraction, model SLO, model latency, model cost, gateway observability, regional routing, model version pinning. Produces route configs, fallback policies, tenant rate limit policies, observability event schemas, cost dashboard specs, BYOK custody plans, gateway SLO sheets.

Use when building or evolving an internal developer platform (IDP), designing paved roads and golden paths, shipping an internal CLI, scaffolding new services, wiring a service catalog (Backstage, Port, Roadie), standing up ephemeral preview environments per PR, building a developer portal, running user research with internal engineers, measuring time to first deploy / DX NPS / adoption, or coaching a team on product mindset for internal tooling. Triggers: platform, internal platform, IDP, internal developer platform, paved road, golden path, dev experience, DX, developer productivity, developer portal, service catalog, Backstage, Port, Roadie, internal CLI, scaffolding, cookiecutter, yeoman, dev container, devbox, ephemeral environment, preview environment, onboarding, time to first deploy, T2FD, self service. Produces internal platform PRDs, paved road designs, CLI command specs, scaffolding templates, catalog entry shapes, ephemeral env specs, adoption dashboards.

Use when designing a test strategy for a feature or service, writing or reviewing unit / integration / e2e / contract / property tests, building test infrastructure (fixtures, factories, test DBs, golden files), investigating a flaky test, raising or lowering coverage where it matters, setting test gates in CI, or planning regression coverage for a release. Triggers: test, testing, QA, coverage, unit test, integration test, e2e, end to end, Playwright, Cypress, Jest, Vitest, pytest, JUnit, mock, stub, fixture, factory, flaky, flake, regression, contract test, property test, fuzz, snapshot. Produces test plans, test code, fixture / factory libraries, flake investigations, CI gate definitions. Not for incident debugging in production, see senior-devops-sre.

Use when designing, building, reviewing, or operating retrieval augmented generation systems: corpus parsing, chunking, embedding, indexing, retrieval (semantic, lexical, hybrid), reranking, citation, evaluation, and ingestion freshness. Covers vector stores (pgvector, Pinecone, Weaviate, Qdrant, Vespa, Milvus, Elastic kNN), embedding models (text-embedding-3, bge-large, nomic-embed, voyage, cohere), BM25 and reciprocal rank fusion, cross encoder rerankers, ColBERT, MMR, and retrieval specific evaluation. Triggers: RAG, retrieval augmented generation, retrieval, embedding, vector, vector store, pgvector, Pinecone, Weaviate, Qdrant, Vespa, Milvus, hybrid search, BM25, lexical, semantic, reranker, cross encoder, ColBERT, MMR, citation, source, chunking, splitter, recursive splitter, document parsing, OCR for RAG, freshness, retrieval eval, recall, precision, NDCG, hit rate, MRR. Produces parsing plans, chunking configs, vector store schemas, hybrid retrieval pipelines, retrieval eval harnesses.

Use when designing, building, evaluating, or operating production ranking and recommendation systems: feed ranking, product recommendations, search ranking, content discovery, ads relevance, related items, you may also like, up next, home feed. Covers two stage retrieval plus ranking, two tower embedding retrieval, learning to rank (LTR), multi objective optimization (relevance plus engagement plus business value), diversity and MMR, exploration vs exploitation, contextual bandits, off policy evaluation (IPS, doubly robust), position bias correction, cold start strategies, and slice based monitoring. Triggers: recommender, recommendation, ranking, feed, search ranking, learning to rank, LTR, two tower, embedding retrieval, candidate generation, multi objective, MMR, diversity, exploration, exploitation, contextual bandit, multi armed bandit, off policy evaluation, IPS, doubly robust, propensity, click through rate, CTR, watch time, engagement, recommender eval, NDCG, MRR, hit rate, recall at k.

Use when designing, building, evaluating, or operating production conversational voice systems: IVR, voice agents, voice assistants, agent voice fronts, voice cloning compliant products. Covers streaming STT (ASR), streaming TTS, real time pipelines over WebRTC and telephony (SIP, Twilio, Vonage, Telnyx), turn taking, barge in detection, VAD (voice activity detection), prosody and SSML, dialog state, latency budgets (time to first audio, end to end response time), telephony codecs (mu law, a law, narrow band, 8 kHz), accessibility (captions, text alternative). Triggers: voice AI, voice agent, STT, speech to text, ASR, TTS, text to speech, voice cloning, WebRTC, telephony, SIP, Twilio, Vonage, Telnyx, dialog state, turn taking, barge in, VAD, prosody, SSML, real time API, streaming TTS, streaming STT, time to first audio, whisper, deepgram, ElevenLabs, Cartesia, Resemble, gpt-4o-realtime. Produces voice latency budgets, barge in specs, telephony integration plans, dialog state schemas, voice eval sets.

Use when writing, reviewing, or upgrading a C# / .NET application on .NET 8 or .NET 9 (with awareness of the .NET 10 LTS release on the horizon). Covers ASP.NET Core minimal APIs and MVC, Entity Framework Core 9, dependency injection, `IOptions<T>` configuration, structured logging with `ILogger<T>` and Serilog, OpenTelemetry, Polly resilience, MediatR, FluentValidation, xUnit plus WebApplicationFactory plus Testcontainers, source generators, Native AOT, modern C# (records, primary constructors, pattern matching, file-scoped namespaces, collection expressions, `IAsyncEnumerable`, channels, `Span<T>`). Triggers: C#, csharp, .NET, dotnet, .NET 8, .NET 9, ASP.NET Core, minimal API, record, primary constructor, pattern match, EF Core, Entity Framework, NuGet, Serilog, OpenTelemetry, AOT, native AOT, Blazor, MAUI, xUnit, NUnit, Moq, Polly, MediatR, source generator, `IAsyncEnumerable`, channels, `Span`, `Memory`. Produces minimal API endpoints, EF Core DbContexts, migrations, hosted services.

Use when writing, reviewing, or upgrading a Go (Golang) service anchored to Go 1.22+ (generics, range over int, log/slog, http.ServeMux method routing). Covers idiomatic error wrapping with fmt.Errorf and errors.Is / errors.As, context.Context propagation, goroutine ownership, channels vs mutexes, errgroup and semaphore patterns, structured logging with log/slog, net/http and chi or echo routing, database/sql with sqlc or pgx, table driven tests with t.Run, the race detector in CI, and pprof profiling. Triggers: Go, Golang, go.mod, go.sum, goroutine, channel, context.Context, ctx, slog, errors.Is, errors.As, defer, panic, recover, sync, mutex, RWMutex, atomic, generics, type parameter, interface, struct, embedding, GOMAXPROCS, race detector, pprof, gRPC Go, net/http, database/sql, sqlx, pgx, sqlc, gorm, gin, chi, echo, fiber. Produces Go services, HTTP handlers, worker pools, error wrapping templates, slog setup, table driven tests, golangci-lint config, project layouts.

Use when writing, reviewing, or upgrading a Java service on JDK 21 LTS or newer (Java 21, 23, 24); designing records, sealed interfaces, pattern matching, switch expressions, text blocks, virtual threads (Project Loom), and structured concurrency; building Spring Boot 3 services (Spring Web, Spring Data JPA, Spring Security, Spring Boot Actuator) or evaluating Quarkus, Micronaut, Helidon for native image; tuning the JVM (G1, ZGC, heap sizing), reading JFR recordings, and running async-profiler; managing Maven (pom.xml) or Gradle (build.gradle.kts) builds; testing with JUnit 5, AssertJ, Mockito, and Testcontainers against real Postgres or Kafka. Triggers: Java, JDK, JVM, Java 17, Java 21, Java 23, Spring, Spring Boot, Spring Boot 3, Spring Web, Spring Data, Hibernate, JPA, record, sealed, pattern matching, switch expression, virtual thread, Project Loom, structured concurrency, Maven, Gradle, pom.xml, build.gradle, GraalVM, native image, Quarkus, Micronaut, JUnit 5, AssertJ, Mockito, Testcontainers, G1.

Use when writing, reviewing, or debugging modern Python (3.12, 3.13) across web services, data pipelines, ML glue, and scripts. Covers type hints (mypy, pyright, PEP 695, `Self`, `Protocol`, `TypedDict`, `Annotated`), packaging with `pyproject.toml`, `uv` installs and lockfiles, ruff lint and format, dataclasses with `frozen=True` and `slots=True`, `pydantic` models, `asyncio` and `TaskGroup`, `asyncio.to_thread` for blocking work, GIL realities and `multiprocessing`, pytest fixtures and parametrize, hypothesis, FastAPI plus pydantic, httpx, sqlalchemy, pandas, polars, numpy, profiling with cProfile, py-spy, scalene, native extensions via Cython or Rust (PyO3). Triggers: Python, Python 3.12, Python 3.13, type hint, mypy, pyright, pyproject.toml, uv, pip, poetry, hatch, ruff, black, dataclass, pydantic, asyncio, async, await, GIL, gevent, multiprocessing, pytest, pytest-xdist, hypothesis, FastAPI, starlette, httpx, requests, sqlalchemy, pandas, polars, numpy, packaging, wheel. Produces `pyproject.toml`.

Use when writing, reviewing, or upgrading a Ruby on Rails app (Rails 7 or Rails 8); designing ActiveRecord models and queries, ActionPack controllers, Hotwire (Turbo, Stimulus, Turbo Streams) views, Sidekiq or Solid Queue workers, Action Mailer, Devise plus Pundit auth, RSpec plus factory_bot tests, and reversible migrations with strong_migrations. Covers N+1 elimination, includes vs preload vs eager_load, scopes, polymorphic and STI tradeoffs, concerns vs service objects, schema.rb vs structure.sql, propshaft and importmaps, Solid Cache, Solid Cable, Russian doll caching, and Rails 7 to Rails 8 upgrade mechanics. Triggers: Rails, Ruby on Rails, ActiveRecord, ActionPack, Hotwire, Turbo, Stimulus, Sidekiq, Solid Queue, GoodJob, Devise, Pundit, strong_migrations, rspec-rails, factory_bot, schema.rb, structure.sql, has_many, polymorphic, concern, strong params, propshaft, importmaps. Produces models, migrations, controllers, Turbo Stream views, jobs, Pundit policies, request specs, upgrade checklists.

Use when building, reviewing, debugging, or shipping React Native apps to the App Store and Play Store. Covers the New Architecture (Fabric renderer, TurboModules, JSI, codegen), Expo and EAS workflows, Expo Router and React Navigation, state with Zustand, Jotai, Redux Toolkit and Tanstack Query, Reanimated 3 worklets, Skia rendering, FlashList, Hermes engine, platform specific code (`.ios.tsx`, `.android.tsx`), native modules in Swift and Kotlin, autolinking, iOS pods, Android Gradle, Maestro and Detox, EAS Build, EAS Submit, and EAS Update. Triggers: React Native, RN, Expo, EAS, Expo Router, Metro, Hermes, New Architecture, Fabric, TurboModule, JSI, codegen, React Navigation, Reanimated, Skia, Zustand, Jotai, Tanstack Query, native module, autolinking, Maestro, Detox, Flipper, EAS Update, `app.json`. Produces RN app skeletons, navigation graphs, state and data layers, animation worklets, native module specs, and EAS pipelines. Not for native-vs-RN-vs-Flutter platform choice, see `senior-mobile-engineer`.

Use when writing, reviewing, or debugging Rust code (stable 1.80 plus, edition 2021 and 2024); designing async services on tokio, axum or actix handlers, tower middleware, sqlx or sea-orm data layers, serde models, tracing plus tracing-subscriber observability, and CLI binaries with clap. Covers ownership and borrowing, lifetimes, traits with associated types and GATs, Send and Sync bounds, pinning, async-trait status, cancellation with tokio::select, error design with thiserror and anyhow or eyre, unsafe contracts, FFI, no_std, and WASM with wasm-bindgen. Triggers: Rust, cargo, Cargo.toml, ownership, borrow, lifetime, trait, async, await, tokio, futures, Send, Sync, Pin, axum, actix, warp, rocket, sqlx, sea-orm, serde, anyhow, thiserror, eyre, tracing, clippy, rustfmt, rust-analyzer, unsafe, FFI, no_std, WASM, wasm-bindgen, clap. Produces error enums, axum services, tracing setups, sqlx queries, cancellation patterns, release profile config, clippy and CI lint config.

Use when building, reviewing, or debugging Tailwind CSS interfaces, design tokens, utility classes, theming, dark mode, container queries, or component libraries built on Tailwind. Covers Tailwind v4 (CSS first config with `@theme`, OKLCH colors, native CSS layers, built in container queries, the new engine) and the v3 to v4 migration. Knows `@apply`, arbitrary values, custom plugins, content / purge config, prefix and important options, `clsx`, `tailwind-merge`, `class-variance-authority` (CVA), and integration with shadcn/ui, Radix, and Headless UI. Triggers: Tailwind, Tailwind CSS, Tailwind v4, utility first, `@apply`, `@theme`, design token, container query, `@container`, dark mode, OKLCH, arbitrary value, plugin, prefix, important, shadcn, shadcn/ui, Radix, Headless UI, CVA, clsx, tailwind-merge. Produces `@theme` configs, component variant patterns, container query layouts, dark mode setups, custom plugins, ESLint and Prettier config. Not for visual design from a blank page, see `senior-ux-designer`.