Browse agent skills

TypeScript SDK for Kanyun Sandbox v2

High-performance, extensible HTTP client for undici

TypeScript HTTP client for the [Marfa](https://marfa.so) API.

Sandboxed bash shell for LLM agents — pure JS, virtual or contained filesystem, structural containment with introspection and resource budgets

Claude Code skill — delegate coding tasks to Opencode, supervise via git diff & save 50-90% Claude token usage

TypeScript HTTP client for Agent Tick.

OpenAI Responses API phase executor for Coro — stateless conversation replay, MCP-backed tools, and provider-owned cost accounting.

Kakunin compliance tools for the Vercel AI SDK — verify agent certificates, check scope, read risk scores, emit audit events

n8n node to invoke Pandoc for document conversion

Cryptographically secure password generator with policies

Generate Lorem Ipsum and placeholder text in various styles

Encode, decode, and verify JWT tokens from terminal

Local-first engineering workflow skills for Codex.

Headless React utilities to send the current page as a prompt to ChatGPT, Perplexity, Claude, and other AI providers. No UI included — bring your own dialog.

Manage .env files: validate, merge, encrypt, generate examples

Run an agent prompt repeatedly in a batch loop, tee-ing each run to the console and a per-run log file.

R'a Core standalone terminal AI coding assistant

Site fetching runtime for agent skills, with Playwright login state and browser/request helpers.

Tool-call event renderers — ToolCallChip / ToolGroup / dedicated chips for write / bash / search / email / excel / pdf / docx / ask-user.

An Autonomous AI Software Engineer in Your Terminal

Cost-optimized model routing for Oh-My-Pi — routes prompts to cheap/mid/expensive models based on task complexity. Tracks per-turn and session costs. Uses TOON compression for conversation history optimization.

Embedding-based retrieval, hybrid scoring, RAG pipeline, and intelligence primitives for @forgrit/design-engine. Optional local LLM and ONNX runtime.

Editorial AI · 直笔 — A Claude Skill for serious writers. Skeleton-first writing, anti-fabrication, diagnosis+prescription, word count discipline. Built for journalists, analysts, PR teams, B2B writers.

Spec Driven Test — the test layer for SDD

Drop-in workflow scaffold for AI-assisted coding — works with Claude Code, Codex, Gemini, and any agent loading the open agent skills standard. Project-agnostic: web apps, backends, mobile, CLIs, libraries. Includes /workflow-init slash command, /feature

Personal Claude Code skills — learning-mode (learn-from-doc, learn-from-zero, learn-by-case) + PR review (address-review). Installs via bunx.

Facet Terminal payment-rail adapter for Stripe (Cards, ACH, wallets). Implements FacetPaymentRailAdapter on top of the official stripe npm package. PaymentIntents with reserve-capture, refunds, disputes, and webhook signature verification.

Workflow control plane for Zoo Code.

CLI for installing and managing Claude Code harness capabilities

Devcontainer scaffolding CLI with per-project AI credential isolation