All authors
Claude Skills by Amey-Thakur
github.com/Amey-Thakur1,001 skills16 installs1,473 views
- Repository HygieneKeep a repository free of stale branches, dead files, oversized objects, and abandoned configuration. Use when a repository has grown cluttered or slow to clone.Votes: 0GitHub stars: 7
- Repository PermissionsGrant repository and organisation access through teams and least privilege, and review it before it drifts. Use when managing access for a growing organisation or auditing who can do what.Votes: 0GitHub stars: 7
- Repository StructureOrganise a repository so a newcomer finds what they need and automation has predictable paths. Use when starting a repository or when nobody can find anything in an existing one.Votes: 0GitHub stars: 7
- Secret ScanningDetect committed credentials, respond correctly when one is found, and prevent the next one. Use when setting up a repository or after a credential appears in history.Votes: 0GitHub stars: 7
- Ai Datacenter NetworkingLay out the network for distributed training so collectives run on the fastest link that spans them, using NVLink, InfiniBand, and topology-aware placement. Use when a training job spans multiple GPUs or nodes and interconnect, not compute, is capping throughput.Votes: 0GitHub stars: 7
- Checkpointing Large TrainingCheckpoint multi-node training runs so a save costs seconds instead of minutes and a resume reproduces the run exactly. Use when a job is large enough that a crash without a recent, verified checkpoint means losing hours of GPU time.Votes: 0GitHub stars: 7
- Cuda Kernel BasicsWrite first CUDA kernels that map work onto the grid, coalesce global memory, and keep enough warps resident to hide latency. Use when hand-writing or reviewing a CUDA kernel and it runs far below the bandwidth or FLOPs the card should reach.Votes: 0GitHub stars: 7
- Distributed Training ScalingScale training from one GPU to many by moving through data parallel and sharded FSDP modes, overlapping communication with compute, and reading the efficiency curve to find the ceiling. Use when adding GPUs stops making training proportionally faster and you need to locate where the scaling leaks.Votes: 0GitHub stars: 7
- Fault Tolerant TrainingKeep a long training job alive across GPU failures, node evictions, and stragglers so one bad host costs minutes, not the whole run. Use when a run spans enough GPUs and hours that hardware failure during the job is expected, not hypothetical.Votes: 0GitHub stars: 7
- Gpu Cluster SchedulingSchedule GPU jobs on a shared cluster so distributed jobs get all their GPUs at once, fragmentation stays low, and preemption is predictable. Use when several teams share a GPU pool and jobs sit pending while GPUs sit idle.Votes: 0GitHub stars: 7
- Gpu Cost PlanningPlan GPU spend by comparing owned hardware, reserved cloud, and on-demand against real utilization and lead time, so you commit at the right break-even. Use when deciding whether to buy GPUs, reserve cloud capacity, or burst on-demand for a workload.Votes: 0GitHub stars: 7
- Gpu Memory HierarchyPlace data across the GPU memory tiers so a kernel is limited by math rather than by repeated trips to HBM. Use when a kernel is memory-bound, when deciding what to stage in shared memory or L2, or when register spills are stalling a hot loop.Votes: 0GitHub stars: 7
- Gpu Sharing MigShare one physical GPU across several small workloads using MIG, MPS, or time slicing, and pick the mode whose isolation matches the risk. Use when GPUs sit at low utilization because each job needs far less than a whole A100 or H100.Votes: 0GitHub stars: 7
- Gpu Utilization MonitoringMonitor a GPU fleet so you can tell a genuinely busy GPU from one reporting 100 percent utilization while computing almost nothing, and find the wasted spend. Use when GPUs look busy on the dashboard but throughput or cost per token says otherwise.Votes: 0GitHub stars: 7
- Inference Serving OptimizationTune LLM serving to hold latency SLOs while raising GPU throughput, working the batch scheduler, KV cache, and paged attention together. Use when a serving replica misses its latency target or leaves memory and utilization on the table.Votes: 0GitHub stars: 7
- Kernel Profiling NsightProfile GPU work with Nsight Systems and Nsight Compute to read the timeline, name the bottleneck class, and pull the metric that dictates the fix. Use when a GPU program is slower than expected and you need evidence before touching a kernel.Votes: 0GitHub stars: 7
- Mixed Precision DeploymentShip FP16, BF16, or FP8 training and inference that holds accuracy while capturing the speedup, using loss scaling and numeric validation. Use when moving a model off FP32 to run faster or fit in less memory and the result must stay correct.Votes: 0GitHub stars: 7
- Model ParallelismSplit a model that will not fit on one GPU across devices with tensor, pipeline, and expert parallelism, choosing each split by its communication cost. Use when weights or activations exceed a single card and you must shard without stalling on the interconnect.Votes: 0GitHub stars: 7
- Onnx Export PipelinesExport a trained model to ONNX that runs portably across runtimes, closing operator gaps and proving numeric parity against the source framework. Use when a PyTorch or TensorFlow model must run outside its training stack and the export must be trusted, not just produced.Votes: 0GitHub stars: 7
- Quantization DeploymentQuantize a trained model to INT8 or INT4 for inference, calibrate the ranges, and gate the release on a measured quality regression. Use when serving needs lower latency and memory and you will spend effort keeping accuracy inside a defined budget.Votes: 0GitHub stars: 7
- Tensor Core UtilizationGet matrix multiplies onto the tensor cores by fixing shapes, precision, and alignment, then measure that the cores actually fired. Use when a GEMM or attention kernel runs far below the card's advertised throughput and you suspect it fell back to the CUDA cores.Votes: 0GitHub stars: 7
- Tensorrt OptimizationCompile a trained model into a fast, GPU-specific TensorRT engine by controlling precision, defining dynamic shape profiles, and proving the fused engine kept its accuracy. Use when a PyTorch or ONNX model must reach a hardware latency floor that eager execution cannot.Votes: 0GitHub stars: 7
- Triton Inference ServerDeploy models on NVIDIA Triton with a valid model repository, ensemble pipelines, and concurrent execution tuned to keep the GPU saturated. Use when serving one or more models through Triton and configuring layout, dynamic batching, instance groups, or a preprocess-plus-inference pipeline.Votes: 0GitHub stars: 7
- Vllm ServingDeploy an LLM on vLLM with continuous batching, deliberate VRAM planning, and multi-model hosting that never overcommits the card. Use when serving on vLLM and deciding memory fraction, context length, tensor parallel size, and how many models share a GPU.Votes: 0GitHub stars: 7
- Character EncodingHandle Unicode end to end so text survives storage, transport, and display without mojibake or truncation mid-character. Use when text arrives corrupted, lengths behave oddly, or emoji break a field.Votes: 0GitHub stars: 7
- Currency LocalizationDisplay, store, and reason about money across currencies without rounding errors or implied conversions. Use when showing prices in more than one currency or storing monetary amounts.Votes: 0GitHub stars: 7
- Locale Aware SortingSort and compare text using locale collation rather than byte order, so lists read correctly in every language. Use when displaying sorted names, searching case-insensitively, or matching user input.Votes: 0GitHub stars: 7
- Locale FallbackDefine what a user sees when a string, a locale, or a region is not available, so gaps degrade predictably instead of showing keys or blanks. Use when supporting partial translations or regional variants.Votes: 0GitHub stars: 7
- Locale FormattingFormat numbers, dates, currency, and units by locale using the platform formatter rather than string templates. Use when displaying any value whose written form differs between regions.Votes: 0GitHub stars: 7
- Plural And Gender RulesHandle plurals, grammatical gender, and agreement using locale plural categories rather than an if-else on count. Use when a string contains a number or refers to a person or object with gender.Votes: 0GitHub stars: 7
- Pseudo LocalizationTest localisation readiness with generated pseudo-translations that expand, accent, and bracket text, before any real translation exists. Use when preparing a product for translation and wanting to find breakage early.Votes: 0GitHub stars: 7
- Rtl LayoutSupport right-to-left languages by mirroring layout, icons, and interactions while leaving numbers and code untouched. Use when adding Arabic, Hebrew, Persian, or Urdu support to an interface.Votes: 0GitHub stars: 7
- String ExternalizationMove user-facing text out of code into catalogs with stable keys and context, so translation becomes possible without touching logic. Use when preparing a product for translation or fixing hardcoded strings.Votes: 0GitHub stars: 7
- Text Expansion LayoutDesign interfaces that survive translated text growing or shrinking substantially, without truncation or broken layout. Use when building UI that will be translated, or fixing clipped text in another language.Votes: 0GitHub stars: 7
- Translation Quality ReviewReview translated text for accuracy, register, and fit in context, with a rubric rather than an impression. Use when accepting translations or diagnosing why a localised product feels wrong to native speakers.Votes: 0GitHub stars: 7
- Translation WorkflowRun translation as a pipeline with catalogs, context, review, and continuous updates rather than a one-off handoff. Use when shipping in multiple languages on an ongoing release cadence.Votes: 0GitHub stars: 7
- Js Async PatternsCompose async JavaScript with promises and async/await correctly, propagating errors and cancelling with AbortController. Use when writing async code, running work concurrently, or debugging swallowed errors and unhandled rejections.Votes: 0GitHub stars: 7
- Js Error HandlingHandle errors in JavaScript and TypeScript with proper Error subclasses, cause chains, and a policy for async and unhandled failures. Use when designing error handling or debugging lost stack traces and swallowed errors.Votes: 0GitHub stars: 7
- Js Event LoopReason about the JavaScript event loop, microtasks vs macrotasks, and why blocking it freezes everything. Use when debugging ordering surprises, UI jank, or code that runs in an unexpected sequence.Votes: 0GitHub stars: 7
- Js ImmutabilityWork with immutable data in JavaScript through structural updates, readonly types, and freeze where it earns its cost. Use when managing shared or reactive state, or debugging bugs from unexpected mutation.Votes: 0GitHub stars: 7
- Js ModulesNavigate ESM and CommonJS, interop between them, and structure imports so bundlers can tree-shake. Use when hitting module-resolution errors, mixing ESM and CJS, or shrinking a bundle.Votes: 0GitHub stars: 7
- Js Tooling SelectionChoose the JavaScript/TypeScript bundler, test runner, and linter for a project by its type, not by fashion. Use when setting up a toolchain or deciding whether to migrate an existing one.Votes: 0GitHub stars: 7
- Monorepo WorkspacesRun a JavaScript/TypeScript monorepo with workspaces, task orchestration, and internal package versioning that scales. Use when managing multiple packages in one repo or when a growing monorepo's builds and installs get slow.Votes: 0GitHub stars: 7
- Node Backend SetupBootstrap a Node.js backend with ESM, typed config, graceful shutdown, and structured logging from the start. Use when starting a Node service or hardening one that grew without foundations.Votes: 0GitHub stars: 7
- Npm PublishingPublish an npm package that installs and imports cleanly: correct exports map, dual formats, types, and semver. Use when releasing a library to npm or fixing a package consumers cannot import.Votes: 0GitHub stars: 7
- Ts Api TypesType the boundaries of a TypeScript codebase: public API surfaces, branded types, and runtime validation of untyped input. Use when designing types others consume, or when external data enters your program.Votes: 0GitHub stars: 7
- Tsconfig MasteryConfigure tsconfig.json deliberately: the flags that matter, module and target settings, and build vs typecheck configs. Use when setting up TypeScript compilation or debugging confusing module and output errors.Votes: 0GitHub stars: 7
- Typescript GenericsWrite TypeScript generics that infer well and stay readable, and know when a generic is not worth it. Use when designing typed reusable functions or APIs, or untangling generic signatures nobody can read.Votes: 0GitHub stars: 7
- Typescript NarrowingNarrow union types safely with guards, discriminated unions, and exhaustiveness checks. Use when working with values that could be several types, or when the compiler will not let you access a property you know is there.Votes: 0GitHub stars: 7
- Typescript StrictnessTurn on and migrate to TypeScript's strict flags incrementally so the compiler catches real bugs. Use when configuring TypeScript strictness or tightening a loose codebase without a big-bang rewrite.Votes: 0GitHub stars: 7