Jason Lo

Search and retrieve preprints from arXiv via the Atom API. Use this skill when searching for papers in physics, mathematics, computer science, quantitative biology, quantitative finance, statistics, electrical engineering, or economics by keywords, authors, arXiv IDs, date ranges, or categories.

Generate CITATION.cff files for GitHub repositories. Creates machine-readable citation metadata that GitHub renders as a "Cite this repository" button with APA and BibTeX export. Use when the user wants to make their repo citable, add a citation file, create a CITATION.cff, or enable the GitHub cite button. Triggers on "citation", "CITATION.cff", "cite this repo", "make citable", "add citation", "how to cite".

Build apps with the Claude API or Anthropic SDK. TRIGGER when: code imports `anthropic`/`@anthropic-ai/sdk`/`claude_agent_sdk`, or user asks to use Claude API, Anthropic SDKs, or Agent SDK. DO NOT TRIGGER when: code imports `openai`/other AI SDK, general programming, or ML/data-science tasks.

Generate VS Code tasks.json from Python CLI commands discovered in a repo (typer, click, argparse, or plain functions). Parses CLI modules to find commands, asks the user which to include, and writes .vscode/tasks.json. Use when the user wants to create VS Code tasks from their CLI, generate tasks.json, or convert CLI commands to VS Code tasks. Triggers on "generate tasks", "vscode tasks", "tasks.json", "cli to tasks", "cli2task", "create tasks from cli".

Optimizes database queries and improves performance across PostgreSQL and MySQL systems. Use when investigating slow queries, analyzing execution plans, or optimizing database performance. Invoke for index design, query rewrites, configuration tuning, partitioning strategies, lock contention resolution.

PostgreSQL table design reference: data types, constraints, indexes, JSONB patterns, partitioning, and best practices. Use when designing PostgreSQL tables, schemas, data models, choosing data types or indexes, creating upsert-heavy or update-heavy tables, or understanding PostgreSQL-specific gotchas. Triggers on PostgreSQL schema, table design, PRIMARY KEY, FOREIGN KEY, indexes, B-tree, GIN, JSONB, constraints, normalization, partitioning, row-level security.

Creates Dockerfiles, configures CI/CD pipelines, writes Kubernetes manifests, and generates Terraform/Pulumi infrastructure templates. Handles deployment automation, GitOps configuration, incident response runbooks, and internal developer platform tooling. Use when setting up CI/CD pipelines, containerizing applications, managing infrastructure as code, deploying to Kubernetes clusters, configuring cloud platforms, automating releases, or responding to production incidents. Invoke for pipelines, Docker, Kubernetes, GitOps, Terraform, GitHub Actions, on-call, or platform engineering.

Check repo state and update all documentation to reflect reality: README.md, docs/ directory, .env.example, CONTRIBUTING.md, CHANGELOG, and other doc files. Use when the user says "update docs", "sync docs", "refresh readme", "docs are stale", or after adding features, refactoring, changing dependencies, or reorganizing project structure. Triggers on "update docs", "sync docs", "update readme", "refresh readme", "docs are stale", "readme is stale", "update documentation", "fix docs".

Use when building high-performance async Python APIs with FastAPI and Pydantic V2. Invoke to create REST endpoints, define Pydantic models, implement authentication flows, set up async SQLAlchemy database operations, add JWT authentication, build WebSocket endpoints, or generate OpenAPI documentation. Trigger terms: FastAPI, Pydantic, async Python, Python API, REST API Python, SQLAlchemy async, JWT authentication, OpenAPI, Swagger Python.

Create production-ready FastAPI projects with async patterns, dependency injection, and comprehensive error handling. Use when scaffolding new FastAPI applications, setting up project structure, implementing repository/service patterns, or bootstrapping backend API projects. Trigger terms: FastAPI scaffold, FastAPI project template, FastAPI boilerplate, new FastAPI app, FastAPI project setup.

Use when fine-tuning LLMs, training custom models, or adapting foundation models for specific tasks. Invoke for configuring LoRA/QLoRA adapters, preparing JSONL training datasets, setting hyperparameters for fine-tuning runs, adapter training, transfer learning, finetuning with Hugging Face PEFT, OpenAI fine-tuning, instruction tuning, RLHF, DPO, or quantizing and deploying fine-tuned models. Trigger terms include: LoRA, QLoRA, PEFT, finetuning, fine-tuning, adapter tuning, LLM training, model training, custom model.

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.

Create stunning, animation-rich HTML presentations from scratch or by converting PowerPoint files. Use when the user wants to build a presentation, convert a PPT/PPTX to web, or create slides for a talk/pitch. Helps non-designers discover their aesthetic through visual exploration rather than abstract choices.

Configure Python projects with modern tooling: uv, ruff, ty. Use when creating new Python projects, setting up pyproject.toml, migrating from pip/Poetry/mypy/black, or writing standalone scripts. Triggers on new Python project, setup Python, pyproject.toml, migrate from pip, migrate from Poetry, replace mypy, replace black, uv init, ruff config.

Convert Python scripts into production-ready Docker and Apptainer/Singularity containers with Pixi dependency management, automatic detection, multi-stage builds, and best practices. Supports GPU/CUDA configurations, multi-environment setups (CPU/GPU), HTCondor/SLURM integration, and .sif conversion for HPC. Use when users need to containerize Python applications, create Dockerfiles, generate Apptainer images, package Python code for deployment, need container configuration for Python projects, or want GPU-accelerated applications for HPC environments. Supports Pixi and pip workflows. Handles scripts, web apps, workers, services, and ML/data science workloads on Docker, Apptainer, and HPC systems.

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.

Discover trending Claude Code skills, study their implementations, and evolve this repo by adopting valuable patterns. Runs in a loop until stopped. Use when the user wants to improve the skills collection, adopt community best practices, or keep the repo current. Triggers on "evolve", "self-evolve", "improve skills repo", "adopt new patterns", "update skills from community".

Create a new versioned release: bump version, tag, and push to trigger the release workflow. Use when the user says "release", "new version", "bump version", "ship it", "cut a release", or "publish".

Audit a repository for unnecessary complexity, dead code, outdated dependencies, and stale TODOs, then propose concrete simplifications. Use when the repo feels bloated, after a big refactor, or to find dead code and overengineering. Triggers on "simplify", "clean up this repo", "reduce complexity", "find dead code", "audit complexity", "overengineered", "too complex", "code health", "tech debt".

Check for skill-sommelier plugin updates and apply them. Compares installed plugin git HEAD against remote, shows changelog, and pulls updates. Use when the user says "update skills", "check for updates", "am I up to date", "update skill-sommelier", or "new skills available".

Research-before-coding workflow. Search for existing tools, libraries, packages, MCP servers, and skills before writing custom code. Use when starting a new feature, adding a dependency, or about to write a utility that likely already exists. Triggers on "add X functionality", "implement Y", "build a Z" — any time custom code is about to be written for a common problem.

Identify similar or overlapping skills in this repo and merge them into one coherent skill. Reduces maintenance burden by eliminating redundancy across skills that share trigger phrases, functionality, or domain. Use when the skill collection feels bloated, two skills seem to do the same thing, or after adding many skills and wanting to deduplicate. Supports a quick mode (trigger overlap only) and full mode (deep analysis). Triggers on "consolidate skills", "merge skills", "deduplicate skills", "combine skills", "overlapping skills", "too many similar skills", "reduce skill count", "skill overlap", "quick overlap check".

Create, improve, and design Claude Code skills. Routes between three modes: create (new skill from scratch), improve (fix quality issues in existing skills), and design (architect multi-step workflow skills). Use when users want to create a skill, turn a workflow into a skill, write a SKILL.md, improve skill quality, fix frontmatter, audit skills, design workflow architecture, or structure multi-step skills. Triggers on "create a skill", "make a skill", "turn this into a skill", "new skill", "fix my skill", "improve skill quality", "skill review", "audit skills", "design a workflow skill", "skill architecture", "multi-step skill".

Search GitHub for trending Claude Code skills, present a personalized ranked table, and install selections. Use when the user wants to find new skills, browse what's available, explore the skills ecosystem, or bootstrap their skill collection. Also serves as the "init" / bootstrap command — when called with no arguments, auto-generates personalized search queries from the user's profile. Triggers on "find skills", "discover skills", "search for skills", "what skills exist", "browse skills", "trending skills", "new skills", "init", "bootstrap skills", "setup skills", "recommended skills", "get started with skills", "what skills should I have".

Self-improving skill optimization using the Karpathy autoresearch pattern. Runs a skill repeatedly, evaluates outputs against binary criteria, mutates the SKILL.md to keep winners, and loops until convergence. Use when the user wants to tune a skill, optimize a skill, improve skill quality with evals, auto-tune a prompt, run autoresearch, benchmark a skill, or self-improve a skill. Triggers on "tune skill", "skill tune", "autoresearch", "optimize skill", "auto-tune", "eval loop", "self-improving", "benchmark skill", "run evals on skill".

Validate all skills in this repo for frontmatter correctness, naming conventions, and structural rules. Use when adding a new skill, before releases, or in CI. Triggers on "validate skills", "lint skills", "check skills", "audit frontmatter", "skill validation", "pre-release check".

Automated weekly skill discovery for GitHub Actions. Uses claude-code-action to search GitHub for new Claude Code skills, filter against installed skills and user profile, and create a GitHub issue with recommendations. Triggers on "weekly discover", "automated discovery", "skill recommendations".

Analyze Claude Code user history to build a rich profile: interests, tech stack, work patterns, preferences, and personality traits. Use when the user wants to understand their coding habits, generate a developer profile, or review how they use Claude Code. Also used by discover-skills for personalized ranking. Triggers on "my profile", "who am I", "analyze my usage", "developer profile", "coding habits", "how do I use Claude".

Suite of tools for creating elaborate, multi-component claude.ai HTML artifacts using modern frontend web technologies (React, Tailwind CSS, shadcn/ui). Use for complex artifacts requiring state management, routing, or shadcn/ui components - not for simple single-file HTML/JSX artifacts.