Loading...
Loading...
Found 106 Skills
How to benchmark and analyze memory usage in Turso using the memory-benchmark crate and dhat heap profiler. Use this skill whenever the user mentions memory usage, memory profiling, allocation tracking, heap analysis, memory regression, memory benchmarking, dhat, or wants to understand where memory is being allocated during SQL workloads. Also use when investigating memory growth in WAL or MVCC mode. IMPORTANT - If you modify the perf/memory crate (add profiles, change CLI flags, change output format, etc.), update this skill document to reflect those changes so it stays accurate for future agents.
[Hyper] Optimize an existing Codex skill through baseline-first experiments, binary evals, optional guards, and one-mutation-at-a-time iteration. Use for skill autoresearch, measured trigger/workflow improvement, self-optimizing a skill, benchmarking skill changes, or resuming skill experiment artifacts.
Optimize existing Triton kernels for NVIDIA TileIR backend on Blackwell GPUs (sm_100+). Adds TileIR-specific autotune configs: occupancy, num_ctas, TMA descriptors. Covers kernel classification (dot-related, norm-like, elementwise, reduction), type-specific transformations, and PTX-vs-TileIR benchmarking. Triggered by: "optimize for TileIR", "add TileIR configs", "Blackwell optimization", "TMA descriptors", "2CTA mode", "occupancy tuning". Kernels use standard `import triton`; TileIR activates via ENABLE_TILE=1 when nvtriton is installed.
Generate the competitive analysis section with competitor profiles, SWOT analysis, competitive matrix, differentiation strategy, market share positioning, and sustainable competitive advantage (moat). Proves the business can win against alternatives. Use when building or reviewing competitive analysis sections, benchmarking against competitors, or defining market positioning. Incorporates Farris's competitive metrics, guerrilla positioning strategy, value-based differentiation frameworks, Teece's business model vs strategy distinction (business model = architecture of value creation and capture; strategy = how the model is made difficult to imitate), Kaza's four differentiation types (aesthetic experience, social experience, boundary interactions, purposeful experiences), Ohmae's 3C Strategic Triangle and Key Factors for Success, and the Portable MBA onstage/backstage model with Value Net complementors framework.
Social listening and brand monitoring strategy — monitoring, Boolean queries, sentiment, competitive intel, crisis detection, AI visibility monitoring, LLM brand mentions. Platform comparison (Meltwater, Brandwatch, Talkwalker, Brand24, Sprout Social, Mention, Hootsuite, BrandJet, Influencity), monitoring setup (keywords, sources, alerts), sentiment analysis, competitive benchmarking (share of voice), crisis detection (real-time alerts, escalation), consumer insights, and reporting. Use when you don't know what people are saying about your brand, competitors are getting mentioned more than you, negative sentiment is spiking and you need to understand why, you're missing PR crises until it's too late, you can't tell if your brand shows up in AI/LLM answers, or you need to pick the right social listening tool. Do NOT use for platform-specific config (use /sales-meltwater), influencer discovery (use /sales-influencer-marketing), social media publishing/scheduling, or SEO keyword research (use /sales-semrush).
Design, build, and optimize dashboards for RIA practice management with AUM tracking, revenue analytics, and KPI frameworks. Use when the user asks about tracking firm-level metrics, monitoring advisor productivity, measuring organic growth rate, analyzing client retention and attrition, building executive or branch manager views, setting up exception alerts for NIGO or rebalancing drift, benchmarking against industry peers, or designing role-based dashboard access. Also trigger when users mention 'how is the practice doing', 'revenue per advisor', 'client attrition', 'net new assets', 'effective fee rate', 'practice benchmarking', 'AUM growth decomposition', 'advisor capacity', or 'referral tracking'.
Automates benchmark test creation for C++ projects using Google Benchmark with consistent software testing patterns. Use when creating performance benchmarks, profiling tests, or when the user mentions benchmarking, Google Benchmark, or performance testing.
Benchmark vLLM or OpenAI-compatible serving endpoints using vllm bench serve. Supports multiple datasets (random, sharegpt, sonnet, HF), backends (openai, openai-chat, vllm-pooling, embeddings), throughput/latency testing with request-rate control, and result saving. Use when benchmarking LLM serving performance, measuring TTFT/TPOT, or load testing inference APIs.
Build institutional-grade comparable company analyses with operating metrics, valuation multiples, and statistical benchmarking in Excel/spreadsheet format. **Perfect for:** - Public company valuation (M&A, investment analysis) - Benchmarking performance vs. industry peers - Pricing IPOs or funding rounds - Identifying valuation outliers (over/under-valued) - Supporting investment committee presentations - Creating sector overview reports **Not ideal for:** - Private companies without comparable public peers - Highly diversified conglomerates - Distressed/bankrupt companies - Pre-revenue startups - Companies with unique business models
Use this skill when the user is writing, debugging, profiling, refactoring, reviewing, benchmarking, parallelising, exporting, or explaining JAX code, or when they mention JAX, jax.numpy, jit, grad, value_and_grad, vmap, scan, lax, random keys, pytrees, jax.Array, sharding, Mesh, PartitionSpec, NamedSharding, pmap, shard_map, Pallas, XLA, StableHLO, checkify, profiler, or the JAX repo. It helps turn NumPy or PyTorch-style code into pure functional JAX, fix tracer/control-flow/shape/PRNG bugs, remove recompiles and host-device syncs, choose transforms and sharding strategies, inspect jaxpr/lowering/IR, and benchmark compiled code correctly.
Generate realistic synthetic evaluation datasets by analyzing the user's codebase, prompts, production traces, and reference materials. Interactive, consultant-style — asks clarifying questions, proposes a plan, generates a preview for approval, then delivers a complete dataset uploaded to LangWatch. Use when user asks to generate, create, or build a dataset for evaluation, testing, or benchmarking.
End-to-end SGLang SOTA performance workflow. Use when a user names an LLM model and wants SGLang to match or beat the best observed vLLM and TensorRT-LLM serving performance by searching each framework's best deployment command, benchmarking them fairly, profiling SGLang if it is slower, identifying kernel/overlap/fusion bottlenecks, patching SGLang code, and revalidating with real model runs.