2,102 件 ・ 関連度順
概要と使いどころ
Serves LLMs with high throughput using vLLM's PagedAttention and continuous batching. Use when deploying production LLM APIs, optimizing inference latency/throughput, or serving models with limited GPU memory. Supports OpenAI-compatible endpoints, quantization (GPTQ/AWQ/FP8), and tensor parallelism.
日本語の概要は準備中です。原文の説明を表示しています。
Runs LLM inference on CPU, Apple Silicon, and consumer GPUs without NVIDIA hardware. Use for edge deployment, M1/M2/M3 Macs, AMD/Intel GPUs, or when CUDA is unavailable. Supports GGUF quantization (1.5-8 bit) for reduced memory and 4-10× speedup vs PyTorch on CPU.
日本語の概要は準備中です。原文の説明を表示しています。
Adds PyTorch FSDP2 (fully_shard) to training scripts with correct init, sharding, mixed precision/offload config, and distributed checkpointing. Use when models exceed single-GPU memory or when you need DTensor-based sharding with DeviceMesh.
日本語の概要は準備中です。原文の説明を表示しています。
Language-independent tokenizer treating text as raw Unicode. Supports BPE and Unigram algorithms. Fast (50k sentences/sec), lightweight (6MB memory), deterministic vocabulary. Used by T5, ALBERT, XLNet, mBART. Train on raw text without pre-tokenization. Use when you need multilingual support, CJK languages, or reproducible tokenization.
日本語の概要は準備中です。原文の説明を表示しています。
Half-Quadratic Quantization for LLMs without calibration data. Use when quantizing models to 4/3/2-bit precision without needing calibration datasets, for fast quantization workflows, or when deploying with vLLM or HuggingFace Transformers.
日本語の概要は準備中です。原文の説明を表示しています。
Use when writing Spark jobs, debugging performance issues, or configuring cluster settings for Apache Spark applications, distributed data processing pipelines, or big data workloads. Invoke to write DataFrame transformations, optimize Spark SQL queries, implement RDD pipelines, tune shuffle operations, configure executor memory, process .parquet files, handle data partitioning, or build structured streaming analytics.
日本語の概要は準備中です。原文の説明を表示しています。
Configures monitoring systems, implements structured logging pipelines, creates Prometheus/Grafana dashboards, defines alerting rules, and instruments distributed tracing. Implements Prometheus/Grafana stacks, conducts load testing, performs application profiling, and plans infrastructure capacity. Use when setting up application monitoring, adding observability to services, debugging production issues with logs/metrics/traces, running load tests with k6 or Artillery, profiling CPU/memory bottlenecks, or forecasting capacity needs.
日本語の概要は準備中です。原文の説明を表示しています。
Writes, optimizes, and debugs C++ applications using modern C++20/23 features, template metaprogramming, and high-performance systems techniques. Use when building or refactoring C++ code requiring concepts, ranges, coroutines, SIMD optimization, or careful memory management — or when addressing performance bottlenecks, concurrency issues, and build system configuration with CMake.
日本語の概要は準備中です。原文の説明を表示しています。
Pick the right serving container for a SageMaker model deployment and find its current image URI. Use this skill whenever about to deploy a model to a SageMaker endpoint and an image URI needs to be chosen — including when the user says "deploy this LLM", "host this HuggingFace model", "serve this fine-tuned model", "deploy this embedding model", "host a reranker", "serve a sentence-transformers model", or when about to hardcode any container URI in deployment code. HuggingFace-curated Deep Learning Containers are ALWAYS preferred: HuggingFace vLLM (LLMs and generative rerankers), HuggingFace vLLM-Omni (multimodal), TEI (embeddings/cross-encoder rerankers), HF Inference Toolkit (other transformers). Generic images (AWS vLLM, DJL-LMI, SGLang) are used only when no HuggingFace image is compatible — never merely because they carry a newer version. Never hardcode a container URI from memory and never default to TGI. Prevents stale-image failures and wrong-region URIs.
日本語の概要は準備中です。原文の説明を表示しています。
Hugging Face CLI to estimate the required memory to load Safetensors or GGUF model weights for inference from the Hugging Face Hub
日本語の概要は準備中です。原文の説明を表示しています。
Jurisdiction-aware wage/hour and employment Q&A — classification, overtime, meal/rest breaks, leave, final pay — answered for the specific state/country with the controlling rule researched and cited rather than stated from memory. Use when the user asks any employment law question, or says "what's the rule in [state]", "is this exempt", "do we have to pay overtime for", or "can we classify this as".
日本語の概要は準備中です。原文の説明を表示しています。
Save the decisions, facts, dead ends and architecture this conversation established to Atlas's shared memory, so other agents and future sessions can build on them.
日本語の概要は準備中です。原文の説明を表示しています。
Pull version-current Convex docs for the version this project uses — pin the installed version, fetch page-as-markdown or check node_modules types, freshness hierarchy — instead of writing a possibly-stale API from memory.
日本語の概要は準備中です。原文の説明を表示しています。
Adapt this C/C++ ASAN vulnerability pipeline to a different vulnerability class, target shape, language, or detection mechanism. Use when the user wants to port, migrate, retarget, customize, or fork the pipeline for something other than C/C++ memory-safety bugs — web apps, smart contracts, deserialization, ML systems, or any other domain.
日本語の概要は準備中です。原文の説明を表示しています。
Sets up and runs Atheris, the coverage-guided Python fuzzer built on libFuzzer. Covers TestOneInput harnesses, FuzzedDataProvider, instrumenting both pure Python and native C extensions, and running under AddressSanitizer. Use when fuzzing a Python package, hunting memory corruption in a Python C extension, or choosing between Atheris and Hypothesis for a Python target.
日本語の概要は準備中です。原文の説明を表示しています。
Sets up and runs Ruzzy, Trail of Bits' coverage-guided Ruby fuzzer and the only production-ready one for the language. Covers harness structure, fuzzing pure Ruby and the native C extensions in gems, and sanitizer builds. Use when fuzzing a Ruby library or gem, testing a Ruby C extension for memory safety, or asking how to fuzz Ruby at all.
日本語の概要は準備中です。原文の説明を表示しています。
Performs comprehensive Rust security review for safe/unsafe boundary issues, memory safety in unsafe blocks, concurrency hazards, panic-induced DoS, FFI safety, and async runtime mistakes. Use when auditing Rust crates, services, or libraries — particularly those with `unsafe`, FFI, or concurrent code.
日本語の概要は準備中です。原文の説明を表示しています。
Comprehensive methodology for auditing and exploiting TLS/SSL implementations and misconfigurations across network services and mobile applications. Covers protocol downgrade attacks including POODLE (CVE-2014-3566) against SSLv3 CBC padding, DROWN (CVE-2016-0800) cross-protocol attack leveraging SSLv2 export ciphers to decrypt TLS sessions, and FREAK (CVE-2015-0204) forcing RSA export-grade key exchange. Addresses BEAST (CVE-2011-3389) exploiting CBC IV predictability in TLS 1.0, CRIME (CVE-2012-4929) and BREACH targeting TLS-level and HTTP-level compression oracles respectively, and Heartbleed (CVE-2014-0160) for OpenSSL memory disclosure. Covers certificate validation bypass techniques for applications with improper hostname verification or chain validation, certificate pinning bypass using Frida and Objection for mobile application interception, HSTS bypass via NTP manipulation and subdomain exploitation, TLS 1.3 0-RTT replay attacks against non-idempotent endpoints, mutual TLS (mTLS) authentication attacks including client certificate theft and relay, and Certificate Transparency log monitoring for reconnaissance. Primary tooling includes testssl.sh for comprehensive TLS auditing, sslyze for Python-integrated scanning, sslscan for quick cipher enumeration, and tlsx for high-speed TLS probing at scale. Maps to CWE-295 (Improper Certificate Validation), CWE-319 (Cleartext Transmission of Sensitive Information), and CWE-757 (Selection of Less-Secure Algorithm During Negotiation).
日本語の概要は準備中です。原文の説明を表示しています。
Anti-forensics and evidence destruction techniques for red team operators conducting authorized engagements. Covers log clearing on Windows (wevtutil, Clear-EventLog, ETW provider patching) and Linux (journal truncation, utmp/wtmp binary editing, syslog manipulation), timestamp manipulation via Timestomp and SetMACE to defeat timeline analysis, filesystem-level anti-forensics including NTFS Alternate Data Streams for payload hiding and secure deletion with sdelete/shred, memory artifact removal to counter live forensics, disk artifact manipulation targeting MFT entries and USN journal records, network forensics evasion through encrypted C2 channels and DNS-over-HTTPS tunneling, and anti-VM/sandbox detection to avoid dynamic analysis environments. Tools: Timestomp, wevtutil, sdelete, shred, MimiPenguin, Invoke-Phant0m. Aligns to MITRE ATT&CK T1070 (Indicator Removal), T1027 (Obfuscated Files or Information), T1497 (Virtualization/Sandbox Evasion). Each technique includes the forensic artifact it targets, the destruction or manipulation method, and the defender perspective so operators understand detection gaps they must account for.
日本語の概要は準備中です。原文の説明を表示しています。
Command and Control framework deployment, configuration, and operational tradecraft for red team engagements. Covers Cobalt Strike (malleable C2 profiles, Beacon types HTTP/HTTPS/DNS/SMB, Beacon Object Files for in-memory execution, sleep and jitter tuning, named pipe pivoting), Sliver (implant generation across mTLS/WireGuard/DNS transport, operator multiplayer mode, armory extensions), Mythic (agent ecosystem with Apollo/Poseidon/Medusa, C2 profile configuration, translation containers), Havoc (Demon agent with sleep obfuscation via Ekko/Zilean, indirect syscalls, dotnet inline execution), Metasploit (msfvenom payload generation, multi/handler staging, Meterpreter post-exploitation modules), redirector architecture using Apache mod_rewrite and Nginx, domain fronting through CDN providers, DNS-based C2 for restrictive network egress, and TLS certificate management for infrastructure OPSEC. Tools: Cobalt Strike, Sliver, Mythic, Havoc, Metasploit Framework. Aligns to MITRE ATT&CK T1071 (Application Layer Protocol), T1573 (Encrypted Channel), T1090 (Proxy/Connection Proxy).
日本語の概要は準備中です。原文の説明を表示しています。
Top-level workflow skill for USD performance diagnosis and optimization. Use for slow loading, high memory, low FPS, or 'optimize my scene' requests; delegates auth/runtime setup to Phase 0 owners.
日本語の概要は準備中です。原文の説明を表示しています。
Capture and inspect iOS leaks and memgraphs. Use when debugging leaked objects, retain cycles, memory growth, or before/after leak evidence.
日本語の概要は準備中です。原文の説明を表示しています。
Build agents and multi-agent systems with the Swarms framework — the Agent class, tools, autonomous loops, memory, and the 15+ multi-agent architectures (SequentialWorkflow, ConcurrentWorkflow, GraphWorkflow, HierarchicalSwarm, SwarmRouter, and more). Use whenever writing, reviewing, or debugging code that imports `swarms`.
日本語の概要は準備中です。原文の説明を表示しています。
Feature planning and implementation with 4 adaptive phases (Specify, Design, Tasks, Execute). Auto-sizes depth by complexity. Writes testable requirements in EARS notation, atomic tasks, atomic Conventional Commits, and requirement traceability. Ships deterministic Python validation scripts so structural gates are enforced by code, not memory. Features an independent Verifier (author != verifier, evidence-or-zero), a discrimination sensor, a decision log (STATE.md), a test-coverage matrix, and a self-improving lessons layer. Stack-agnostic and tool-agnostic. Use when (1) planning features, (2) implementing with verification and atomic commits, (3) validating an implementation against a spec. Triggers on "specify feature", "discuss feature", "design", "tasks", "implement", "validate", "verify work", "UAT", "record decision", "pause work", "resume work". Do NOT use for pure architecture decomposition analysis or standalone technical design documents.
日本語の概要は準備中です。原文の説明を表示しています。