Browse agent skills

Run scoped, task-specific inference inside web applications.

讓 AI 出一份原創的學測模擬考：命題、解題驗算、套用大考正式版面，交付題目與答案詳解兩份 PDF。支援 Claude、ChatGPT、Gemini；學測七科可用，會考與分科測驗規劃中。

OpenCode TUI sidebar section: context-window usage as a progress bar.

Session-level observability tool for LLM coding agents (opencode/Claude Code)

Empirical AI coding agent benchmarks: Monolith vs Chemical X Architecture

Empirical AI coding agent benchmarks: Monolith vs Chemical X Architecture

Automatic strong/cheap model-tier routing for DeepSeek Harness: intent-gated tier landing on the agent/request waterfall, plan-mode handoff (strong plans, cheap implements), deterministic high-risk guards on tools/pre-execute, failure escalation with TTL

One row per OpenCode tool call. Click to reveal its output, with native file diffs and no exploration grouping.

Local-first AI coding agent token usage dashboard and CLI

Agent-oriented Android and iOS runtime automation through MCP.

HLS.js adapter for OptiView Ads SDK

OptiView Ads SDK entry package - OptiViewAds with a default HLS.js ad player (native fallback)

OptiView Ads SDK Core - Player-agnostic ad insertion

Automation engine for managing search deindexing, content removal requests, review issues, and online reputation workflows across major platforms.

The outer loop of the autonomous SDLC harness: init, doctor, config and daemon management for the harness Claude Code plugin, plus opt-in local docs retrieval.

泛微 e-cology E9 HTTP 能力 CLI 封装（待办、微博、新建流程、邮件、公文、日程、考勤、建模台账）

Agent-to-Map (A2Map) Protocol, Controller and Differential Renderer for MapLibre GL JS

Evaluation definitions, scorers, datasets, and runners for B4.run agents.

CNIPA 中国发明专利申请全流程 Agent Skill：在先专利划界 → 五书撰写 → 300DPI 黑白线条附图 → 数据溯源 → 提交底稿与回归校验 → 提交后通知书接收与法定期限管理。An Agent Skill for drafting and filing invention patents with CNIPA.

Automation flows (flows-as-data engine)

Deterministic harnesses, fixtures, and matchers for testing B4.run agent applications.

Context window analyzer for AI coding agents. Coming soon.

SQL connection, catalog, query, and result workbench for the official DSH right Sidebar.

Claude Skills for adversarial evaluation of ideas, decisions and research papers, plus humaner, a ruleset and scanner for writing without the signs of AI-generated text. Each skill is a portable SKILL.md for Claude.ai and Claude Code.

A loopengine ability: a zero-config, no-API-key search tool over your own project's docs, plus a skill teaching an agent when and how to use it.

Versioned TypeScript and JSON Schema contracts for AIDrift by ZettaCore.

Docker and Kubernetes sandbox providers for isolated B4.run workspace execution.

Framework-agnostic embeddable chat SDK for oyehello agents. Web component + headless client.

An OMC-shaped delivery harness for Meta Muse Code: deep-interview -> ralplan -> ralph, plus deep-dive/trace, team and cancel.

Command-line development, testing, build, and runtime tools for B4.run applications.