Zard Li
Community@zardli1115 · Hangzhou
Best Wishes!
Agent Skills by Zard Li
Showing 168 vetted skills indexed across 2 GitHub repositories.
readme-skill
Generate fact-based GitHub README files from repository evidence and user confirmation.
google_meet
Join Google Meet URLs and transcribe live captions into a transcript.
parallel-cli
Automate web search, extraction, and deep research with structured JSON outputs.
searxng-search
Query a SearXNG instance for aggregated search results with JSON responses.
gitnexus-explorer
Index a git codebase into a navigable knowledge graph with a web UI.
qmd
Search and rerank passages across locally indexed markdown knowledge bases.
drug-discovery
Retrieve bioactivity and molecule property data from public biomedical databases.
duckduckgo-search
Search DuckDuckGo for text, news, images, and videos via ddgs CLI.
scrapling
Fetch static and JavaScript-rendered pages with anti-bot methods and CSS/XPath extraction.
bioinformatics
Routes computational biology tasks to bioSkills references or ClawBio runnable pipelines.
domain-intel
Generate passive domain reconnaissance with subdomains, TLS, DNS, WHOIS, and availability signals.
docker-management
Manage Docker containers, images, volumes, networks, and Compose stacks via CLI.
inference-sh-cli
Search app IDs and run cloud AI apps via the infsh CLI.
watchers
Poll RSS/Atom feeds, GitHub repositories, and JSON endpoints for new items.
sherlock
Search 400+ platforms for accounts matching a username via Sherlock CLI.
oss-forensics
Correlates Git, GitHub, and web archive evidence to produce forensic reports.
1password
Set up the 1Password CLI to read, inject, and run commands with secrets.
rest-graphql-debug
Diagnose REST and GraphQL API failures by isolating connectivity, TLS, authentication, and parsing layers.
fitness-nutrition
Retrieve exercises from wger and nutrition data from USDA FoodData Central.
neuroskill-bci
Connect to a running NeuroSkill instance to interpret real-time BCI cognitive and emotional metrics.
zedclaw-atropos-environments
Builds ZedClaw Atropos RL environments with AgentResult message parsing and ToolContext-based reward validation.
nemo-curator
Filter noisy multimodal datasets and remove exact, fuzzy, and semantic duplicates.
huggingface-accelerate
Convert standard PyTorch training scripts into distributed and mixed-precision workflows.
chroma
Stores embeddings and metadata for filtered, similarity-based retrieval.
Frequently Asked Questions About Zard Li
FAQPage SchemaWhat tasks can I accomplish using Zard Li's skills?▼
You can fine-tune and serve LLMs (PEFT, TRL, vLLM, llama.cpp), run vector search with Qdrant, FAISS, Pinecone, and Chroma, generate Excel financial models (DCF, LBO, merger, comps), create images, video, and music (Stable Diffusion, Manim, AudioCraft), and manage GitHub repositories, pull requests, and code reviews.
Who is the target audience for this skill registry?▼
ML engineers running distributed training with FSDP2 and torchtitan, quantitative finance analysts building three-statement and valuation models, software developers needing debugging and PR-review pipelines, and creative producers generating diagrams, pixel art, infographics, and music from text prompts.
How do the machine learning training skills work in practice?▼
Training skills configure existing frameworks: Axolotl and Unsloth use YAML-driven LoRA/QLoRA setups, TRL covers SFT, DPO, PPO, and GRPO alignment, torchtitan and PyTorch FSDP handle 4D-parallel pretraining, and Hugging Face Accelerate converts standard PyTorch scripts to mixed-precision distributed runs.
What prerequisites and dependencies do these skills require?▼
Dependencies vary per skill: GPU skills need NVIDIA hardware with CUDA for TensorRT-LLM and Flash Attention; finance skills require Excel-compatible .xlsx tooling; GitHub skills need gh CLI or REST API authentication via HTTPS tokens or SSH keys; macOS skills like iMessage and Apple Notes require the imsg and memo CLIs.
Do the skills cover data retrieval and search use cases?▼
Yes. Search skills query SearXNG, DuckDuckGo, arXiv, and Yahoo Finance; vector databases (Qdrant, FAISS, Pinecone, Chroma) provide embedding similarity retrieval; watchers poll RSS/Atom feeds and GitHub repositories; and scrapling fetches JavaScript-rendered pages with CSS/XPath extraction.