Browse agent skills

The Manta design system, knowledge-graph UI and agent components

The Manta GraphQL contract and the typed React Query hooks generated from it

Namespace prefixes, CURIE helpers and JSON-LD types for the Manta knowledge graph

TypeScript types for the Manta knowledge graph, generated from its SHACL shapes with LDkit

RDFox Datalog parser and AST builder with inference verification

Voice agent framework powered by JEV (Joint Embedding Vectors) for dynamic conversation flow

A local fake Doklado for issuing invoices and fetching their PDFs.

Lat knowledge graph validation and navigation without UI or search

Local-first Web console for Claude Code, OpenCode, Codex, and Kimi Code workflows.

论文知识收集库：多源论文检索(arXiv/Semantic Scholar/OpenAlex/Crossref/DBLP/PubMed)·全文获取·深度学习·多视角总结与复现指南·本地知识库沉淀。Paper research knowledge base plugin for DeepSeek Harness: multi-source search, full-text fetching, deep reading, summarize & reproduce, persistent KB.

Fayz SDK — plugin-workforce plugin

Nekutima AI SDK — vendor-neutral LLM Engine access (DeepSeek, OpenAI, Anthropic, Gemini, xAI, OpenRouter, custom) plus the Pro.vi.li platform client (chat, RAG, agents)

Fine-tune a small LLM on your website's own pages (local LoRA/QLoRA via Unsloth on NVIDIA or MLX on Apple Silicon) and embed it as a WebGPU chat window that serves the model tier matching each visitor's hardware. Generic — works with any static site folde

URL-based structured OCR provider runtime with a Mistral adapter and compatible legacy parsers.

Write pi subagents and compose workflows, simply

Command Code API provider for opencode — use Claude, GPT, Gemini, DeepSeek, Qwen, Kimi, GLM, MiniMax, and Step models via Command Code

Model-specific Lyria boundary: prompt translation, transport, and config interpretation.

Scheduled prompts for the pi coding agent: save a prompt + schedule + tool allowlist as a cron job. Runs inside pi on any OS, keeps configuration and run history, catches up missed and interrupted runs.

Agentic coding assistant

Install skills from the SourceWeft marketplace onto a local coding agent.

DSH 插件：面向 AI 漫剧 / AI 微电影 / AI 视频创作的多项目无限画布编辑器 —— 画布节点编排、分门别类的素材库、全局素材搜索与跨项目软引用，配合 dsh-llm-multimodal 的生成工具与 media-preview 的聊天内预览。

Hosted memory engine with server-side embeddings, SEAL encryption, and semantic search via memwal.

On-device MiniLM embeddings (all-MiniLM-L6-v2) with hybrid vector + keyword recall.

Trajectory harvester extension for pi.dev — monitors local worker model runs and intercepts compilation failure loops.

One-command installer for the Website Builder Kit — downloads your licensed copy, installs it, and launches your AI assistant.

Laya typed decisions (noul / choice / score) as a first-class Cordis service and model-visible tools for DeepSeek Harness.

Official TypeScript/JavaScript client for AI Gate — one API for chat and image models across providers with latency-tuned routing and failover

Server-side text-to-speech finishing and caching: deterministic voice-version keys, Piper and ffmpeg argument builders, injected process and file adapters, and a memory cache for tests.

Local GLiNER2.5 context pruning for OpenCode

Terminal UI for Runstack — chat with your toolkits from the terminal via @ai-sdk/tui.