Browse agent skills

Mine fail-to-pass evaluation tasks from your repo's git history and benchmark AI coding-agent harness configurations against them.

An open, framework-agnostic format for what a test run produced — the .evidence pack, and the library + CLI that validate and seal it.

Install and drive the Fortress stealth Chromium engine. Prebuilt binary, no source.

[BETA] Chat com os modelos do GitHub Copilot no navegador — instala a extensão do portal no VS Code e abre tudo pronto

Turns a plain notes repo into a hub for cross-repo work.

Open-source PostgreSQL credit ledger, usage metering, and billing SDK for AI SaaS

Semantic search companion module for kbexplorer — index production and consumption over knowledge graphs

A living project brain and refine-in-place toolbar for Claude Code and Codex projects.

Modelgov API client. `feature` and `userType` are mandatory on every request.

Optimized Antigravity skills and MCP pack for BDB DEV

Terminal-native personal intelligence and coding CLI.

Typed client and codegen CLI for the Valtrix API. Run valtrix generate to get a client typed against your organization's synced records.

Conclave AI Ollama agent — local inference via Ollama's OpenAI-compatible endpoint. Zero cost, zero API key.

Conclave AI Claude agent — one-shot tool-use review via @anthropic-ai/sdk (agent-sdk migration deferred per decision-status #8).

Conclave AI Telegram bot-runner — long-polls Telegram callback_query events and dispatches repository_dispatch actions back to the target repo's GH Actions workflows.

Reusable wrapper around the GitHub Copilot CLI: schema-validated JSON output, timeouts, and a warm/persistent ACP variant that keeps `copilot --acp` alive to drop per-call cold start.

The electronic line — build agents at the speed of light.

EngramX memory plugin for OpenClaw — persistent, decentralized agent memory

React Native PDF viewer with high performance, zoom, and support for large PDF files. Fast, memory-efficient, and supports annotations (annots) on iOS and Android.

A geeky product-engineering team for AI coding agents.

opencode closed-loop usage coach — quota SENSE -> coaching DECIDE -> loop ACT + TUI integration

Drop-in production trace collector for LLM agents (TypeScript/JS). Streams Vercel AI SDK, Anthropic, OpenAI, and LangChain.js traces to your Mirrors backend with negligible overhead.

Make your AI coding agent ship changes you can trust. One-command installer for the Coding Quality Loop skill.

CLI tools and the reference engine for the Structured Semantic Syntax System

FoundRuu CLI - AI開発を標準化するためのプラットフォームCLI

opencode plugin: adds a 'CliProxy' login that routes the native Anthropic provider through a self-hosted CLIProxyAPI endpoint, with a connection check. Pure proxy: no OAuth dependency, the client sends only an api-key and the proxy handles Anthropic auth

MS-RAGS(ALL-IN-ONE) — production-grade RAG framework builder terminal CLI

<!-- Generated file — edit the source area record and regenerate; do not edit by hand. -->

OpenCode plugin that lets text-only models handle image uploads by describing them via a vision model before they reach the LLM.

Restart llama-server with different model configurations. For single-model setups where models need different flags (context size, vision projector, sampling params).