Top Agent Skills

Anthropic Claude API patterns for Python and TypeScript. Covers Messages API, streaming, tool use, vision, extended thinking, batches, prompt caching, and Claude Agent SDK. Use when building applications with the Claude API or Anthropic SDKs.

FOX v2.0 — Dual-mode emerging movers scanner. All live trading lessons from Fox v1.0 applied, plus one experimental tweak: Stalker minReasons = 3. Stalker entries must have at least 3 distinct scoring reasons, not just pass the score threshold. This forces breadth of confirmation beyond the auto-awarded base climb score + one bonus. Stalker minScore 7, minTotalClimb 8, tighter Phase 1 for low-score entries, consecutive-loss streak gate. XYZ banned. Leverage 7-10x. DSL exit managed by plugin runtime via runtime.yaml.

COBRA v1.1 — Arena Sprint Predator. Trades the single most dominant SM asset with maximum conviction. Fee-optimized maker orders. Wider DSL lets positions breathe. Designed to win Arena weeks.

RHINO — Momentum pyramider. Top 10 assets by OI + volume. Enters small (30% of max) on high-conviction convergence, then adds to winners at +10% ROE (40% more) and +20% ROE (final 30%). Thesis re-validated before every add — 4h trend intact, SM aligned, volume present. DSL High Water Mode trails the full position. The only skill in the zoo that builds into winners instead of entering full size and hoping.

🦀 Hyperliquid Claw — OpenClaw Skill Definition

BARRACUDA v1.0.1 — Funding Decay Collector. Finds assets where extreme funding has persisted for 6+ hours, confirmed by SM alignment from Hyperfeed and 4H trend structure. Enters to collect funding while riding the trend. Double edge: price appreciation + funding income from trapped traders on the other side. Fixes Croc's -42.7% failure by requiring SM and trend confirmation before entering.

This skill should be used when building prediction market integrations, implementing betting functionality, or querying Polymarket odds. Covers market search, odds checking, and position management.

Checks a GitHub, GitLab, or Perforce (p4) pull request (or merge request, or shelved changelist) for unresolved review comments, failing status checks, and incomplete PR descriptions. Waits for pending checks to complete, categorizes issues as actionable or informational, and optionally fixes and resolves them. Use when the user wants to check a PR/MR/CL, address review feedback, or prepare a change for submission.

Autonomous run-loop for the blog pipeline. Picks the top unwritten keyword from the vetted queue, dispatches /blog-pipeline in autonomous mode, auto-publishes on success, quarantines persistent failures, and refreshes the keyword queue when empty. Cron-friendly. No human in the loop.

Foundation skill for all RoadRunner workflows: MATLAB path setup, connection, project/scene/scenario lifecycle, world settings, handle management, status, and close. Use when connecting to RoadRunner, managing projects/scenes/scenarios, setting world origin, checking status, closing RoadRunner, or when any downstream RoadRunner skill needs initialization.

Import raw data (CSV, XLSX, TXT, or MATLAB tables) into formats used by Sensor Fusion and Tracking Toolbox. Handles both ground truth trajectories and sensor detection data. For truth: builds trackingScenarioRecording, tuning timetable, truthlog, or converted table. For sensor data: builds task-oriented dataFormat structs (preferred) or objectDetection arrays (legacy). Use when importing flight logs, GPS logs, radar detections, IR measurements, lidar/camera bounding boxes, ADS-B data, AIS ship tracks, or any recorded data for use with trackers, filter tuning, or tracker evaluation.

Share MATLAB content by guiding users through uploading to GitHub, MATLAB Drive, or File Exchange, then generating "Open in MATLAB Online" URLs. Covers the full sharing workflow: choosing a platform, uploading content (automated via gh CLI for GitHub, manual for MATLAB Drive and File Exchange), and constructing the correct URL. Use when a user wants to share MATLAB code with others, open local files in MATLAB Online, generate an open-in-MATLAB-Online button or badge, or when an AI agent has generated MATLAB code locally and the user wants to share it or run it in MATLAB Online.

Build custom PCB structures with pcbComponent, shapes, Boolean ops, feeds, and multi-layer stackups for non-catalog geometries. TRIGGER: user asks to build, modify, or customize a pcbComponent — add/remove shapes, edit polygons, place feeds, add metal layers, cut slots, or create non-catalog RF structures. Also when modifying geometry of an existing catalog-designed component (e.g., adding pads, removing elements, editing vertices). Invoke BEFORE writing pcbComponent code — layer/shape/feed API is non-obvious. SKIP: designing catalog components like filters/couplers/txlines (use the specific matlab-design-pcb-* skill), material/stackup definition only (use matlab-manage-pcb-material), EM analysis (use matlab-analyze-em), importing PCB files (use matlab-read-pcb-layout).

Read BEFORE writing any code that builds or simulates OFDM systems. The default MATLAB patterns for OFDM (e.g., direct ifft/fft, awgn with 0 dBW power, missing symOffset) produce subtly incorrect results — always use ofdmmod/ofdmdemod instead of direct IFFT/FFT. This skill specifies the correct calling conventions for ofdmmod, ofdmdemod, ofdmChannelResponse, ofdmEqualize, and critical anti-patterns that must be avoided. Use when building OFDM transmitters or receivers, allocating subcarriers and guard bands, inserting pilots, computing SNR for OFDM, configuring fading channels (Rayleigh/Rician), estimating and equalizing channels, implementing timing and frequency synchronization, adding LDPC coding, designing resource grids, or setting up OFDM link simulations.

Testing standards using JUnit 5, AssertJ, Mockito, Cucumber, and Spring Boot integration tests for Java. Use when writing or reviewing Java test behavior, including parallel execution and BDD; defer Kotlin-only tests, virtual-thread test infrastructure, and coverage-report/tooling configuration.

Run a read-only SQL query against a SQLite database file

Bulk-import credentials into the Paladin vault from .env files, CSV/JSON exports, Bitwarden, or 1Password — or discover where credentials live. Returns names/kinds/counts only; secret values never enter the agent's context.

Send email via SMTP or SendGrid with subject and body

List pending and completed tasks in the Custodian queue

Check if an HTTP/HTTPS endpoint is reachable and measure latency

Create a new Stripe subscription for a customer

Color design specifications for Naive UI. Invoke when user needs to understand the color palette, main colors, semantic colors, or neutral colors used in Naive UI.

Audit CI/CD pipelines (GitHub Actions) for cost, speed, and safety. Use when the Actions bill is high, Actions minutes, runner cost, slow CI, artifact/cache storage, or "CI/CD / workflow audit". Gate logic (bypass, ratchet gaming, required-but-not) → audit-gate-logic.

Split current work into small reviewable PRs. Use when the user asks to split a chat, set of changes, branch, or PR.

Managed vector database for production AI applications. Fully managed, auto-scaling, with hybrid search (dense + sparse), metadata filtering, and namespaces. Low latency (<100ms p95). Use for production RAG, recommendation systems, or semantic search at scale. Best for serverless, managed infrastructure.

Activation-aware weight quantization for 4-bit LLM compression with 3x speedup and minimal accuracy loss. Use when deploying large models (7B-70B) on limited GPU memory, when you need faster inference than GPTQ with better accuracy preservation, or for instruction-tuned and multimodal models. MLSys 2024 Best Paper Award winner.

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

Simplest distributed training API. 4 lines to add distributed support to any PyTorch script. Unified API for DeepSpeed/FSDP/Megatron/DDP. Automatic device placement, mixed precision (FP16/BF16/FP8). Interactive config, single launch command. HuggingFace ecosystem standard.

Merge multiple fine-tuned models using mergekit to combine capabilities without retraining. Use when creating specialized models by blending domain-specific expertise (math + coding + chat), improving performance beyond single models, or experimenting rapidly with model variants. Covers SLERP, TIES-Merging, DARE, Task Arithmetic, linear merging, and production deployment strategies.

High-level PyTorch framework with Trainer class, automatic distributed training (DDP/FSDP/DeepSpeed), callbacks system, and minimal boilerplate. Scales from laptop to supercomputer with same code. Use when you want clean training loops with built-in best practices.

---

Framework for building LLM-powered applications with agents, chains, and RAG. Supports multiple providers (OpenAI, Anthropic, Google), 500+ integrations, ReAct agents, tool calling, memory management, and vector store retrieval. Use for building chatbots, question-answering systems, autonomous agents, or RAG applications. Best for rapid prototyping and production deployments.

---

Expert guidance for GRPO/RL fine-tuning with TRL for reasoning and task-specific model training

Merge multiple fine-tuned models using mergekit to combine capabilities without retraining. Use when creating specialized models by blending domain-specific expertise (math + coding + chat), improving performance beyond single models, or experimenting rapidly with model variants. Covers SLERP, TIES-Merging, DARE, Task Arithmetic, linear merging, and production deployment strategies.

Compress large language models using knowledge distillation from teacher to student models. Use when deploying smaller models with retained performance, transferring GPT-4 capabilities to open-source models, or reducing inference costs. Covers temperature scaling, soft targets, reverse KLD, logit distillation, and MiniLLM training strategies.

Post-training 4-bit quantization for LLMs with minimal accuracy loss. Use for deploying large models (70B, 405B) on consumer GPUs, when you need 4× memory reduction with <2% perplexity degradation, or for faster inference (3-4× speedup) vs FP16. Integrates with transformers and PEFT for QLoRA fine-tuning.

Use when the user asks to generate, remix, poll, list, download, or delete Sora videos via OpenAI’s video API using the bundled CLI (`scripts/sora.py`), including requests like “generate AI video,” “Sora,” “video remix,” “download video/thumbnail/spritesheet,” and batch video generation; requires `OPENAI_API_KEY` and Sora API access.

Cloud laboratory platform for automated protein testing and validation. Use when designing proteins and needing experimental validation including binding assays, expression testing, thermostability measurements, enzyme activity assays, or protein sequence optimization. Also use for submitting experiments via API, tracking experiment status, downloading results, optimizing protein sequences for better expression using computational tools (NetSolP, SoluProt, SolubleMPNN, ESM), or managing protein design workflows with wet-lab validation.

Creates detailed, sectionized implementation plans through research, stakeholder interviews, and multi-LLM review. Use when planning features that need thorough pre-implementation analysis.

Mass spectrometry analysis. Process mzML/MGF/MSP, spectral similarity (cosine, modified cosine), metadata harmonization, compound ID, for metabolomics and MS data processing.

Build, audit, debug, and migrate websites for durable organic discovery and reliable use by search engines, LLMs, and browser agents. Use for SEO audits or implementation; new sites, routes, templates, CMSs, catalogs, or content programs; crawling, rendering, indexing, robots, canonicals, redirects, sitemaps, hreflang, structured data, metadata, links, faceted navigation, JavaScript SEO, Core Web Vitals, or migrations; keyword, intent, architecture, content, local, media, or authority work; traffic, ranking, CTR, or indexation losses; AEO, GEO, AI Overviews, AI Mode, ChatGPT search, crawler controls, agent-readable pages, Markdown representations, `Accept: text/markdown` content negotiation, `llms.txt`, WebMCP, or agent interaction design. Trigger whenever code or architecture can materially affect public discoverability or machine usability, even if the user does not say SEO. Do not use for paid-search management alone.

Use when the user wants to negotiate a price, rate, compensation, contract, purchase, hotel, service, vendor deal, discount, upgrade, fee, renewal, refund, or other commercial terms; asks what to say to improve leverage or get a better deal; provides an offer/counteroffer and wants a response; or wants to maximize coupons, direct-booking value, bundles, competing configurations, or all-in economics. Do not trigger for ordinary writing, simple price lookup/comparison, arithmetic, generic persuasion, or conflict mediation without a concrete negotiated outcome.

Trigger: REST API, GraphQL schemas, HTTP methods, RFC 7807, endpoint structure, API contracts. Scope: API schema design, error models, request/response structures. Boundary: Excludes client-side visual layouts.

Trigger: marketing copy, landing text, CTA labels, value proposition, copywriting copy, conversion copy. Scope: Persuasive copy generation, conversion-focused taglines. Boundary: Excludes technical specifications or backend engineering.

Upgrade gstack to the latest version.

AI agent that consolidates extracted sales data into live reporting dashboards with territory, rep, and pipeline summaries

Expert data engineer specializing in building reliable data pipelines, lakehouse architectures, and scalable data infrastructure. Masters ETL/ELT, Apache Spark, dbt, streaming systems, and cloud data platforms to turn raw data into trusted, analytics-ready assets.

Design and improve revenue operations, lead lifecycle rules, scoring, routing, handoffs, and CRM process automation. Use when marketing, sales, and customer success workflows need clearer operational structure.

Crypto wallet, token swaps, cross-chain bridges, and access to paid external services (search, image gen, web scraping, AI, and more) via x402 payments.

Guidance on reading, parsing, and processing various file formats (CSV, JSON, YAML) commonly used in legal and compliance tasks

Route security, compliance, privacy, crypto/Web3, framework security, vulnerability review, and risk prompts. Use when prompts mention security-review, security scans, bounty hunting, HIPAA, PHI, compliance, Web3 risk, smart contracts, x402, LLM trading risk, or framework-specific security.

Scan codebases for technical debt, score severity, track trends, and generate prioritized remediation plans. Use when users mention tech debt, code quality, refactoring priority, debt scoring, cleanup sprints, or code health assessment. Also use for legacy code modernization planning and maintenance cost estimation.

API Design Reviewer

Use when the user asks to generate API tests, create integration test suites, test REST endpoints, or build contract tests.

Extract carrier invoice, shipment references, rates, accessorials, taxes, and total charges from freight bills.

Forem API V1 API skill. Use when working with Forem API V1 for api. Covers 58 endpoints.

Neon API skill. Use when working with Neon for projects, api_keys, consumption_history. Covers 138 endpoints.

The Plaid API skill. Use when working with The Plaid for asset_report, cra, credit. Covers 325 endpoints.

Turso Platform API skill. Use when working with Turso Platform for organizations, locations, auth. Covers 46 endpoints.