深度學習

Resolves ambiguous organism names to precise NCBI taxonomy IDs and scientific names, then searches for genomic data in ENA (European Nucleotide Archive). Use this skill when users provide common names (like "malaria parasite", "E. coli", "mouse"), abbreviated names, or when you need to convert any organism reference to an exact scientific name for API queries. This skill handles disambiguation through conversation and validates taxonomy IDs via NCBI Taxonomy API.

Microsoft Pragmatic Rust Universal Guidelines. Use when reviewing naming conventions, code style, import organization, or applying foundational Rust idioms like Option/Result combinators and destructuring.

Create new Claude Code skills with proper structure, modular organization, and performance optimization. Use when building new skills, refactoring existing skills, or improving skill maintainability.

Use this skill when new users want to get started with LimaCharlie, set up their first organization, or begin collecting security data. Guides beginners through org creation and helps identify what to onboard, then hands off to specialized skills.

Audio signal processing library for PyTorch. Covers feature extraction (spectrograms, mel-scale), waveform manipulation, and GPU-accelerated data augmentation techniques. (torchaudio, melscale, spectrogram, pitchshift, specaugment, waveform, resample)

This skill should be used when users need to rent GPUs, run ML training jobs, or manage compute resources on Basilica's decentralized GPU marketplace. Use it for PyTorch/TensorFlow training, distributed training setup, GPU rental management, cost monitoring, or any Basilica CLI workflows. Includes workaround for non-TTY environments like Claude Code.

Understand and navigate the DevPrep AI 7-folder architecture. Use this skill when asked about code organization, where to place new features, what modules exist, or when starting development tasks that need architecture context. Auto-triggers on keywords like "where should", "add module", "architecture", "structure", "organize", "place code", "what modules".

Explains Linux directory structure compared to Windows, where files live, and how to navigate the filesystem. Use when the user asks about Linux folders, directory structure, where to find files, or mentions confusion about Linux file organization.

Automates complete git workflows including branch management, atomic commits with formatted messages, history cleanup, and PR creation. Use when the user wants to commit/make commits, push to remote, create/open a PR, clean up commits, create branches, write commit messages, mentions atomic commits, git workflow, git best practices, or needs help organizing git changes. Also triggers when user is on main/master with uncommitted changes (suggest branching), has messy commit history to clean up before pushing, wants to squash or reorder commits, or needs help creating pull requests.

Personal GTD-based productivity system called LOPS for managing tasks, projects, and priorities in Linear with AI assistance. Use when triaging inbox items, prioritizing work, running weekly reviews, capturing tasks, clarifying next actions, or when the user mentions GTD, productivity, task management, or organizing their work.

Training models on extended context lengths using optimized RoPE scaling and memory-efficient attention kernels. Triggers: long context, max_seq_length, rope scaling, large context window, flex attention.

Build well-structured React components using variant patterns, pure components, and container patterns. Provides patterns for component organization, type safety, and separation of concerns.Use when: building components with multiple display modes, organizing component hierarchies, implementing container/presentational patterns, reducing component proliferation while maintaining clear APIs.

This skill should be used when developing Electron applications with Feature-Sliced Design (FSD) architecture and React 19. Triggers on requests to create components, features, entities, widgets, or pages following FSD layer structure. Also applies when setting up Electron main/preload/renderer process code organization.

Analyze, diagnose, and reorganize story tree structure. Use when user says "check tree health", "find orphans", "move story", "rename story", "fix tree structure", "reparent stories", "validate tree", or when structural issues are suspected in story-tree.db. Focuses Claude on diagnosis while delegating mechanical operations to deterministic scripts. (project)

Apply consulting methodologies from McKinsey, BCG, Bain, and Accenture for structured problem-solving and strategic analysis. Use when analyzing business problems, developing strategy, structuring presentations, evaluating M&A, sizing markets, improving profitability, managing projects, or driving organizational change.

High-level training framework for PyTorch that abstracts boilerplate while maintaining flexibility. Includes the Trainer, LightningModule, and support for multi-GPU scaling and reproducibility. (lightning, pytorch-lightning, lightningmodule, trainer, callback, ddp, fast_dev_run, seed_everything)

This skill should be used when setting up organization-level app integrations (Reddit, Notion, LinkedIn, WordPress) with OAuth flows, encrypted token storage, API client wrappers, and usage logging in a Next.js App Router application. Use this skill when implementing external provider connections for a multi-tenant app with secure credential management, token refresh, and admin-controlled integration features.

Creates, edits, and manages Obsidian vault content including notes, templates, daily notes, and dataview queries. Use when working with markdown files in an Obsidian vault, creating notes, writing templates, building dataview queries, or organizing knowledge management content.

Exporting PyTorch models to ONNX format for cross-platform deployment. Includes handling dynamic axes, graph optimization in ONNX Runtime, and INT8 model quantization. (onnx, onnxruntime, torch.onnx.export, dynamic_axes, constant-folding, edge-deployment)

Manage component registry for task categorization. List, find, and add components. State stored in .claude/state/task-streams/component-manager.json (gitignored). Auto-initializes with C00 "General / Cross-Cutting" on first use. Returns component codes like C01, C02. Use when user says 'list components', 'add component', 'find component', 'categorize tasks by component', or when organizing tasks by system module. (project, gitignored)