404 skills · sorted by relevance
RBG controller stress test - deploy kwok in existing cluster, run create/update/delete load tests with real controller pod, collect pprof profiling and logs, generate analysis report with recommendations
The English summary is being prepared. Showing the original description.
优化实际模型推理链路，将正确性对齐、分段 profiling、显存与数据搬运、TensorRT/ONNX/PyTorch 后端、attention/kernel、FP8/compile、缓存与少步采样、质量回归、GPU 成本和服务验收串成同一实验闭环。当用户要求推理提速、降低显存或 GPU 成本、复现模型效果、定位 GPU 利用率低、优化图像/视频/扩散模型或自托管 LLM 时使用，提供 model inference performance optimization 的脱敏案例。纯 API 调用、提示词或 token 用量优化，以及电脑卡顿，不进入本技能的模型推理实验流程。
The English summary is being prepared. Showing the original description.
Profile programs at the function/method level to identify performance hotspots, bottlenecks, and optimization opportunities. Records execution time, memory usage, and call frequency for each interval. Generates actionable recommendations and visualizations. Use when users need to (1) analyze program performance, (2) identify slow functions or bottlenecks, (3) optimize execution time or memory usage, (4) profile Python, Java, or C/C++ programs with test cases or workload scenarios, or (5) generate performance reports with flame graphs and recommendations.
The English summary is being prepared. Showing the original description.
Automatically instruments source code to collect runtime information such as function calls, branch decisions, variable values, and execution traces while preserving original program semantics. Use when users need to: (1) Add logging or tracing to code for debugging, (2) Collect runtime execution data for analysis, (3) Monitor function calls and control flow, (4) Track variable values during execution, (5) Generate execution traces for testing or profiling. Supports Python, Java, JavaScript, and C/C++ with configurable instrumentation levels.
The English summary is being prepared. Showing the original description.
Set up the Graphsignal Profiler for inference workloads — vLLM, SGLang, PyTorch, and dstack services. Use when the user wants GPU profiling, tracing, or monitoring for inference, asks about `graphsignal-run` or `graphsignal.watch()`, or asks about CUPTI / Prometheus / OTLP setup.
The English summary is being prepared. Showing the original description.
Iterative measurement-driven improvement loop. Measure, profile, mutate, re-measure, commit. Works for performance, bundle size, complexity, test coverage — anything quantifiable. Use when the user wants to systematically improve a metric through repeated cycles of profiling and targeted changes.
The English summary is being prepared. Showing the original description.
Perform systematic exploratory data analysis to understand dataset structure, distributions, relationships, and anomalies before modeling. Use when a dataset is new, its quality is unknown, or the user requests open-ended profiling; use data-analysis instead for a defined hypothesis or decision question.
The English summary is being prepared. Showing the original description.
Analyze datasets to answer defined questions through statistical methods, trend identification, hypothesis testing, and correlation analysis. Use when the user needs evidence-backed findings or decisions from data; use exploratory-data-analysis instead for open-ended first-pass profiling before questions are defined.
The English summary is being prepared. Showing the original description.
Think and work like an expert High-Performance Computing Specialist. Use when a task calls for High-Performance Computing Specialist judgment. Reasons from NUMA topology and hybrid MPI+OpenMP+CUDA decomposition through Slurm fairshare/backfill job design, strong/weak scaling (Amdahl/Gustafson), Darshan/mpiP/Nsight profiling, and parallel HDF5/MPI-IO on Lustre while treating I/O storms, collectives bottlenecks, and rank-binding mistakes as first-class failure modes.
The English summary is being prepared. Showing the original description.
Think and work like an expert Embedded Systems Engineer. Use when a task calls for Embedded Systems Engineer judgment. Reasons from hardware timing, interrupt latency, memory maps, and deterministic WCET resource bounds through DWT CYCCNT and Saleae logic-analyzer timing proofs, SWD/JTAG (J-Link/OpenOCD) with CFSR/HFSR/BFAR fault decode, FreeRTOS/Zephyr respecting configMAX_SYSCALL_INTERRUPT_PRIORITY, MISRA C:2012 and ISO 26262 ASIL discipline, Nordic PPK2 power profiling, and MCUboot signed OTA while treating priority inversion, Cortex-M7 uncached-DMA D-cache incoherence, CAN bus-off, I2C NACK storms, and brownout-during-flash-write NVS corruption as first-class failure modes.
The English summary is being prepared. Showing the original description.
Think and work like an expert Database Systems Researcher. Use when a task calls for Database Systems Researcher judgment. Reasons from storage hierarchy, concurrency semantics, query-optimization theory, and declared workload models through TPC-C/H, YCSB, and JOB benchmarks, Jepsen/Elle correctness checkers, and perf/blktrace/fio profiling while treating cardinality-estimation plan regressions, tail-latency spikes under skew, unfair fsync-disabled speedups, and benchmark-trick wins as first-class failure modes.
The English summary is being prepared. Showing the original description.
Think and work like an expert Catalysis Engineer. Use when a task calls for Catalysis Engineer judgment. Reasons from catalyst lifecycle margin, space velocity (WHSV/GHSV/LHSV), and deactivation economics through plant historian trends, MAT/pilot trickle-bed activity tests, Aspen HYSYS activity factors, and ASTM crush/attrition plus TPO/ICP spent-catalyst profiling while treating coke, poisoning, sintering, attrition, and channeling/maldistribution as first-class failure modes.
The English summary is being prepared. Showing the original description.
Data-independent acquisition (DIA) proteomics analysis with DIA-NN and other tools. Use when analyzing DIA mass spectrometry data with library-free or library-based workflows for deep proteome profiling.
The English summary is being prepared. Showing the original description.
Analyzes CUT&RUN (Skene Henikoff 2017) and CUT&Tag (Kaya-Okur 2019) chromatin profiling data. Handles SEACR vs MACS2 peak calling (with the btaf375 2025 benchmark guidance), pA-MNase vs pA-Tn5 vs pAG-Tn5 chimera differences, E. coli spike-in carryover normalization, IgG-only control logic (no input), characteristic fragment-size signatures (25-75 bp for CUT&Tag), and lower depth requirements (5M reads typical vs 25M for ChIP). Use when calling peaks from CUT&RUN/CUT&Tag, scaling by E. coli spike-in carryover, choosing SEACR norm mode, or comparing CUT&RUN/Tag results to traditional ChIP.
The English summary is being prepared. Showing the original description.
Identify direct miRNA-target interactions from AGO HITS-CLIP, AGO-CLEAR-CLIP (chimeric reads), HEAP (Halo-Ago2 mouse), chimeric eCLIP / miR-eCLIP (deep miRNA-target profiling), or CLASH using chimeric-read processing pipelines, seed-pairing analysis, and 3' auxiliary pairing rules. Use when distinguishing direct miRNA targets from indirect, integrating CLIP-derived target maps with TargetScan / miRDB / DIANA predictions, applying canonical 7mer-8mer seed matching with 3' UTR context, or recovering miRNA-mRNA chimeras at scale.
The English summary is being prepared. Showing the original description.
Workflow for ribosome profiling, P-site aware preprocessing, periodicity checks, ORF detection, and translation efficiency analysis.
The English summary is being prepared. Showing the original description.
Research companies, professionals, job markets, and industry content on LinkedIn through SandBase. Use when asked for LinkedIn research, company analysis, professional profiling, job market research, or B2B competitive intelligence.
The English summary is being prepared. Showing the original description.
Research WeChat Official Account articles, analyze account profiles, track article performance, and read comments. Covers article discovery, account profiling, engagement statistics, comment analysis, and content strategy research on China's dominant messaging ecosystem.
The English summary is being prepared. Showing the original description.
Search videos, analyze channel profiles, track live streams, and research content engagement on WeChat Channels (Video Accounts). Covers video search, channel profiling, live stream monitoring, comment analysis, and content collections.
The English summary is being prepared. Showing the original description.
Build targeted outreach lists with verified contact information, professional context, and conversation starters. Combines Apollo company data, LinkedIn profiling, and email verification to produce ready-to-send outreach campaigns for sales, PR, and partnership teams.
The English summary is being prepared. Showing the original description.
Research potential leads with verified contact information, professional backgrounds, company context, and social presence. Combines LinkedIn profiling, email verification, company enrichment, and social signals for sales-ready prospect intelligence.
The English summary is being prepared. Showing the original description.
Test native, React Native, hybrid, and Flutter mobile apps with Appium 3.x, Detox, Maestro, and Patrol. Covers device farm setup (BrowserStack, Sauce Labs), gesture simulation, deep link and cold-start testing, push notifications, biometric (Face ID) auth, offline/poor-network simulation, and iOS/Android permission dialog handling. Use when: "mobile test," "Appium," "Detox," "Maestro," "Patrol," "Flutter test," "iOS test," "Android test," "device farm," "deep link," "biometric," "Face ID," "permission dialog," "React Native test." Not for: device/browser matrix strategy in the abstract — use cross-browser-testing; app startup/memory/battery profiling depth — use performance-testing; mobile screenshot diffing — use visual-testing. Related: ci-cd-integration, cross-browser-testing, performance-testing, test-data-management, test-reliability.
The English summary is being prepared. Showing the original description.
Optimize MATLAB code for better performance through vectorization, memory management, and profiling. Use when user requests optimization, mentions slow code, performance issues, speed improvements, or asks to make code faster or more efficient.
The English summary is being prepared. Showing the original description.
Prepare requests for the selected AKA Execution entry point, consult Gateway or SSH guidance, and preserve complete execution output and artifacts. Use for handing prepared GPU commands, evaluation, or profiling work to AKA Execution; experiment planning and result interpretation belong to their callers.
The English summary is being prepared. Showing the original description.