Top Agent Skills

Framework for state-of-the-art sentence, text, and image embeddings. Provides 5000+ pre-trained models for semantic similarity, clustering, and retrieval. Supports multilingual, domain-specific, and multimodal models. Use for generating embeddings for RAG, semantic search, or similarity tasks. Best for production embedding generation.

Meta's 7-8B specialized moderation model for LLM input/output filtering. 6 safety categories - violence/hate, sexual content, weapons, substances, self-harm, criminal planning. 94-95% accuracy. Deploy with vLLM, HuggingFace, Sagemaker. Integrates with NeMo Guardrails.

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.

Control LLM output with regex and grammars, guarantee valid JSON/XML/code generation, enforce structured formats, and build multi-step workflows with Guidance - Microsoft Research's constrained generation framework

State-of-the-art text-to-image generation with Stable Diffusion models via HuggingFace Diffusers. Use when generating images from text prompts, performing image-to-image translation, inpainting, or building custom diffusion pipelines.

GPU-accelerated data curation for LLM training. Supports text/image/video/audio. Features fuzzy deduplication (16× faster), quality filtering (30+ heuristics), semantic deduplication, PII redaction, NSFW detection. Scales across GPUs with RAPIDS. Use for preparing high-quality training datasets, cleaning web data, or deduplicating large corpora.

Optimizes transformer attention with Flash Attention for 2-4x speedup and 10-20x memory reduction. Use when training/running transformers with long sequences (>512 tokens), encountering GPU memory issues with attention, or need faster inference. Supports PyTorch native SDPA, flash-attn library, H100 FP8, and sliding window attention.

Meta's 86M prompt injection and jailbreak detector. Filters malicious prompts and third-party data for LLM apps. 99%+ TPR, <1% FPR. Fast (<2ms GPU). Multilingual (8 languages). Deploy with HuggingFace or batch processing for RAG security.

Extract structured data from LLM responses with Pydantic validation, retry failed extractions automatically, parse complex JSON with type safety, and stream partial results with Instructor - battle-tested structured output library

Distributed training orchestration across clusters. Scales PyTorch/TensorFlow/HuggingFace from laptop to 1000s of nodes. Built-in hyperparameter tuning with Ray Tune, fault tolerance, elastic scaling. Use when training massive models across multiple machines or running distributed hyperparameter sweeps.

Optimizes LLM inference with NVIDIA TensorRT for maximum throughput and lowest latency. Use for production deployment on NVIDIA GPUs (A100/H100), when you need 10-100x faster inference than PyTorch, or for serving models with quantization (FP8/INT4), in-flight batching, and multi-GPU scaling.

Evaluates LLMs across 60+ academic benchmarks (MMLU, HumanEval, GSM8K, TruthfulQA, HellaSwag). Use when benchmarking model quality, comparing models, reporting academic results, or tracking training progress. Industry standard used by EleutherAI, HuggingFace, and major labs. Supports HuggingFace, vLLM, APIs.

Extend context windows of transformer models using RoPE, YaRN, ALiBi, and position interpolation techniques. Use when processing long documents (32k-128k+ tokens), extending pre-trained models beyond original context limits, or implementing efficient positional encodings. Covers rotary embeddings, attention biases, interpolation methods, and extrapolation strategies for LLMs.

Train Mixture of Experts (MoE) models using DeepSpeed or HuggingFace. Use when training large-scale models with limited compute (5× cost reduction vs dense models), implementing sparse architectures like Mixtral 8x7B or DeepSeek-V3, or scaling model capacity without proportional compute increase. Covers MoE architectures, routing mechanisms, load balancing, expert parallelism, and inference optimization.

Write publication-ready ML/AI papers for NeurIPS, ICML, ICLR, ACL, AAAI, COLM. Use when drafting papers from research repos, structuring arguments, verifying citations, or preparing camera-ready submissions. For systems venues (OSDI, NSDI, ASPLOS, SOSP), use systems-paper-writing instead.

Anthropic's method for training harmless AI through self-improvement. Two-phase approach - supervised learning with self-critique/revision, then RLAIF (RL from AI Feedback). Use for safety alignment, reducing harmful outputs without human labels. Powers Claude's safety system.

Trains large language models (2B-462B parameters) using NVIDIA Megatron-Core with advanced parallelism strategies. Use when training models >1B parameters, need maximum GPU efficiency (47% MFU on H100), or require tensor/pipeline/sequence/context/expert parallelism. Production-ready framework used for Nemotron, LLaMA, DeepSeek.

Fast tokenizers optimized for research and production. Rust-based implementation tokenizes 1GB in <20 seconds. Supports BPE, WordPiece, and Unigram algorithms. Train custom vocabularies, track alignments, handle padding/truncation. Integrates seamlessly with transformers. Use when you need high-performance tokenization or custom tokenizer training.

Provides guidance for performing causal interventions on PyTorch models using pyvene's declarative intervention framework. Use when conducting causal tracing, activation patching, interchange intervention training, or testing causal hypotheses about model behavior.

Provides guidance for interpreting and manipulating neural network internals using nnsight with optional NDIF remote execution. Use when needing to run interpretability experiments on massive models (70B+) without local GPU resources, or when working with any PyTorch architecture.

Comprehensive guide for writing systems papers targeting OSDI, SOSP, ASPLOS, NSDI, and EuroSys. Provides paragraph-level structural blueprints, writing patterns, venue-specific checklists, reviewer guidelines, LaTeX templates, and conference deadlines. Use this skill for all systems conference paper writing.

Guides researchers through structured ideation frameworks to discover high-impact research directions. Use when exploring new problem spaces, pivoting between projects, or seeking novel angles on existing work.

Applies cognitive science frameworks for creative thinking to CS and AI research ideation. Use when seeking genuinely novel research directions by leveraging combinatorial creativity, analogical reasoning, constraint manipulation, and other empirically grounded creative strategies.

Use this skill when the user asks about FMCSA Compliance Safety Accountability (CSA) program scoring — the seven BASIC categories, BASIC Severity Indicator (BSI), peer percentiles, intervention thresholds, how violations age out, SMS Methodology, ISS (Inspection Selection System), what "alert" status means, Safety Measurement System mechanics, or DataQ disputes. Cite SMS Methodology v3.20.

Use this skill when the user asks about cargo securement requirements under 49 CFR 393 Subpart I — Working Load Limit (WLL), tiedown counts, tiedown types, commodity-specific securement rules (lumber, metal coils, paper rolls, concrete pipe, intermodal containers, vehicles, large boulders, dressed lumber, building materials), CVSA out-of-service criteria, and how to calculate aggregate WLL. Cite 49 CFR 393.100-136.

Use this skill when the user asks about Hours of Service (HOS) rules under 49 CFR 395 — the 11-hour driving limit, 14-hour on-duty window, 60/70 hour rule, 30-minute break, 10-hour off-duty reset, 34-hour restart, sleeper berth provisions, split sleeper berth, short-haul exemption (150-air-mile), agricultural exemption, adverse driving conditions, ELD malfunction handling, personal conveyance, yard moves, or anything related to driver duty status. Cite specific 49 CFR 395 subsection. If the question involves a non-standard operation (passenger carrier, oilfield, agricultural, military), confirm which set of rules applies.

Use this skill when the user asks about Motor Vehicle Records (MVRs), 49 CFR 391.25 annual MVR review, continuous MVR monitoring services (SambaSafety, Foley, etc.), state MVR fees, what to look for on an MVR, CDLIS (Commercial Driver License Information System), or how often MVRs must be pulled. Cite 49 CFR 391.25 and 391.27.

Use this skill when the user asks about FCRA adverse action procedures — what to do when a background check (MVR, criminal, drug test) returns information that may lead to denying employment or terminating a CDL driver. Covers the 5-business-day pre-adverse-action notice, FCRA Summary of Rights, final adverse action notice, state-specific add-ons (CA, NY, etc.), and how to handle disputes. Cite 15 USC § 1681 (FCRA) and 49 CFR 391.

Use this skill when the user asks about DOT audits, FMCSA compliance reviews, new entrant safety audits (NESA), focused investigations, what auditors actually ask for, audit document checklists, common audit failures, the audit ratings (Satisfactory / Conditional / Unsatisfactory), or how to prepare for an upcoming audit. Cite 49 CFR 385.

Use this skill when the user asks about towing + recovery operations — separate FMCSA authority class, light-duty vs medium-duty vs heavy-duty wrecker rules, AAA + AAA-Plus contracts, Roadside-Assistance (RA) operations, "non-consensual tow" rules + state-by-state, and how towing operations differ from general freight. Reference state Public Utility Commission + 49 CFR 395 application.

Use this skill for transporting logs. Covers § 393.116 specific rules, bunks, stakes, and tiedowns required.

Use this skill when the user asks about cybersecurity for a trucking carrier — protecting ELD data, customer information, financial systems, EDI integrations, fleet telematics, ransomware risk, vendor / shipper data sharing, incident response, and FMCSA cyber-recommendations. Reference NIST + CISA cybersecurity frameworks.

Use this skill when a driver needs to extend the 14-hour duty window using the 16-hour exception under § 395.1(o). Covers eligibility, frequency limits (once per 7 days), return-to-terminal requirement, and documentation.

Use this skill for transporting pipes (water, gas, oil, structural). Covers § 393.122 specific rules and roll-prevention.

Use this skill when the user asks about Electronic Logging Device (ELD) compliance under 49 CFR 395.20-38 — ELD mandate, AOBRD vs ELD transition, ELD exemptions, ELD malfunctions, 8-day paper backup, ELD certification on FMCSA registered list, data file integrity, edits + annotations, driver portability, and how to handle ELD failures at roadside. Cite 49 CFR 395.20-395.38.

Use this skill when your carrier is selected for an IFTA audit. Covers what records to produce, how the audit is conducted, and common findings.

Use this skill when the user asks about FMCSA investigation types — Focused Investigation vs Comprehensive Review vs Compliance Review (CR) vs Investigation, when each is triggered, scope differences, on-site vs off-site, and how to prepare for each. Cite 49 CFR 385 Subpart C + D.

Use this skill when the user asks about driver fitness-for-duty programs — pre-trip readiness assessment, recognizing impairment from drugs/alcohol/medication/fatigue/mental health, refusing assignments when unfit, supervisor training on fitness assessment, FMCSA recommendations on duty status determination, and how to handle a driver who appears unfit. Reference 49 CFR 392.3 + 392.4.

Use this skill when the user asks where to find specific information at fmcsa.dot.gov — the SAFER (Safety and Fitness Electronic Records System), Carrier Snapshot, Compliance Safety Accountability (CSA) data, PSP (Pre-Employment Screening Program), Clearinghouse, Unified Carrier Registration (UCR), MCS-150 update, complaints, DataQs, and the FMCSA Portal login. Reference fmcsa.dot.gov.

Use when a carrier asks about tire program design, when to retread vs replace, position management (steer vs drive vs trailer), inflation pressure standards, casing management, scrub rates, road-call prevention, or vendor program selection (Bridgestone, Michelin, Continental, Goodyear). Covers 49 CFR 393.75 and DOT roadside out-of-service criteria.

Use this skill when the user asks about CMV vehicle inspections under 49 CFR 396 — DVIRs (Driver Vehicle Inspection Reports), pre-trip inspections, post-trip inspections, periodic inspections, annual inspections, who can perform inspections, OOS criteria, and what makes a "no defect" DVIR valid. Cite 49 CFR 396.

Use this skill when the user asks about accident claims management — first-notice-of-loss procedures, working with insurance adjusters, claim severity classification, subrogation, claim closure timelines, and how to influence claim outcomes. Different from accident-investigation-playbook (which covers facts); this covers the insurance + claim side.

Use this skill when the user asks about commercial driver drug & alcohol testing under 49 CFR 382 + 40 — the six testing reasons (pre-employment, random, post-accident, reasonable suspicion, return-to-duty, follow-up), MRO process, SAP/return-to-duty process, post-accident triggers, random selection rates, what panels are tested, refusal definitions, or testing programs and consortiums. Cite specific 49 CFR 382 or 40 subsection.

Use when a carrier asks about fuel-card vendor selection (Comdata, EFS, WEX, RTS, Pilot Flying J, Love's Connect), discount levels, fraud control, driver authorization PINs, IFTA reporting features, cash advance, money codes, or how the card fits into payroll/settlement workflows.

Use this skill when a driver does not meet federal medical standards and needs to apply for a variance. Covers application steps, required documentation, and review timeline.

Use this skill when selecting an MRO for drug testing program oversight. Covers MRO qualifications under Part 40, what the MRO does, and how to evaluate MRO services.

Use this skill when the user asks how to interpret a DOT medical examiner's findings — what restrictions mean (corrective lenses, hearing aid, no insulin, etc.), what conditions trigger short-term vs 24-month certificates, vision/hearing thresholds, blood pressure categories, sleep apnea screening, when to seek an exemption program, and how to handle a downgrade or denial. Cite 49 CFR 391.41-49 and the FMCSA Medical Examiner Handbook.

Use this skill when the user asks about onboarding a new CDL driver — pre-hire process, day-zero requirements, application + DQF setup, road test, drug & alcohol pre-employment, Clearinghouse Full query, MVR initial pull, PSP report, ELDT verification, sequence of activities + timing, and what NOT to do (handing keys before testing complete). Cite 49 CFR 391, 382.

Use this skill when the user asks about Driver Qualification Files (DQF) — what must be in a DQ file under 49 CFR 391, retention periods, what's required pre-hire vs ongoing, document expiration tracking, audit-readiness for driver files, common DQF deficiencies that auditors find, or how to handle missing documents on existing drivers. Always cite the specific subsection of 49 CFR 391.

Use this skill when a driver is being screened for sleep apnea. Covers FMCSA Medical Review Board (MRB) recommendations, BMI thresholds, sleep study referrals, and continued certification.

Use this skill when a crash has been incorrectly classified as DOT-recordable when it shouldn't be. Covers the § 390.5 criteria, common misclassifications, and challenge process.

Use this skill when the user asks about CMV roadside inspection levels (Level I through Level VIII), what each level inspects, average inspection times, CVSA decal eligibility, what triggers a level upgrade, and how to coach drivers through a roadside inspection. Reference CVSA North American Standard Inspection Program.

Use this skill when a carrier's CSA peer group changes. Covers what triggers reassignment, how percentile changes affect alert status, and how to interpret the shift.

Use this skill when re-hiring a driver who has been away from CMV operation for 1+ years. Covers re-qualification, refresh of medical and MVR, and whether prior employer inquiries restart.

Use this skill when the user asks about carrier-driver relationships — Independent Contractor (IC / 1099) vs Employee (W-2), Owner-Operator lease agreements, the AB5 California law, the DOL 2024 IC test, lease-purchase programs from the carrier side, settlement statements, deductions, and how to structure a compliant IC arrangement. Cite 49 CFR 376 + 1099 IRS rules.

Use this skill when the user asks about Environmental Health & Safety (EHS) for trucking — fuel spill response, hazmat release reporting, EPA + state environmental reporting, emergency contact information (CHEMTREC), environmental impairment liability insurance, and how to handle a CMV-related environmental incident. Cite 40 CFR + 49 CFR 174.6.

Use this skill to file your IFTA quarterly fuel tax return. Covers filing windows, calculation, payment, and refund situations.

Use this skill when the user asks about oversize / overweight load permits — state-by-state permitting process, escort vehicle (pilot car) requirements, route surveys, federal axle weight limits (Federal Bridge Formula), holiday restrictions, multi-state permits, and how to determine if a load needs a permit. Reference state DOT permitting offices.

Use this skill when the user asks about passenger-carrying CMVs — buses, motorcoaches, charter operations, school buses, limousines, shuttle services. Covers different HOS rules (10-hour driving / 15-hour on-duty / no 34-restart), Class C CDL with P endorsement, S endorsement for school bus, FMCSA Hours-of-Service for passenger vehicles, $5M insurance minimum, BMC-32 filing, accessibility requirements, and how passenger ops differ from property. Cite 49 CFR 395.5 + 387.33.

Use this skill when the user asks about CMV maintenance program requirements under 49 CFR 396.3 — what records must be kept, retention periods (12-18 months on vehicle file + 1 year after disposal), what counts as systematic inspection/repair/maintenance, preventive maintenance scheduling, brake adjustment records, and what auditors look for in vehicle files. Cite 49 CFR 396.3 + 396.21.