GH·RADAR

See before you delete: a cross-platform disk analyzer and system cleaner (Windows/macOS) with a Rust core, Tauri desktop app and agent-friendly CLI. Inspired by Mole.

Whether Claude Code's Bash sandbox works on this machine, the merged sandbox settings and where each came from, why a command ran inside the sandbox, outside it or not at all, and which sandbox settings can never take effect. Zero tokens.

Visual editor for Claude Code skills (SKILL.md). Paint over a skill with Mario Paint-style stamps and Claude rewrites it more human and more principled; erase, promote and demote with a brush. Edits copies, never your originals. Works with ~/.claude/skills, plugin skills and Agent Skills.

Unofficial Go client for the Val Town API, generated from the official OpenAPI spec.

Back up Claude Code, Cursor and GitHub Copilot chats, memory, skills and settings on Windows to OneDrive. One double-click, or one sentence to your AI agent.

Adversarial 3-tier verification skill (Generator, Polymorphic Critic, Judge) with Zero-Context Blind Audit and deductive anti-sycophancy scoring for AI coding agents.

OmniSeek: Multi-lens codebase discovery, deterministic citation grounding verifier & interactive HTML/SARIF dossier generator (MCP + CLI)

OmniEvolve: Evolutionary code & prompt optimization framework with island-model tournaments, evaluator SDK, and SQLite lineage ledger (MCP + CLI)

JevLab: Offline LLM-as-Judge evaluation, position/verbosity bias diagnostics, Platt calibration & agreement benchmark studio (MCP + CLI + UI)

Run the INITE Triad — self-hosted OAuth/OIDC (Auth) + billing (Billing) + AI-agent memory (Brain) — with one compose file. amd64 + arm64.

Command-line tool that imports Moneta bank statements (XML), sorts every transaction into categories using your own keyword rules, skips accounts you want to ignore (e.g. savings), and prints an income and expense overview. Pure Python standard library, exact Decimal money handling, fully typed and tested.

Universal agent studio, turn any AI into an agent with any API format, Loops, Throttle, Routing Tables, more

Caniuse for MCP servers: check which clients (Claude Code, Codex, ...) silently drop or mangle your tools - a static preflight against a versioned quirks corpus.

Snapshot and restore your iTerm2 hotkey window: tabs, panes, colours and the Claude Code session in each pane

AI automation pipeline that turns a script into a fully stitched short-form video using scene-by-scene generation

Curated, production-tested prompt library for sales outreach, AI support bots, structured output, and content generation

Connect your own model providers to Claude Desktop. Open-source macOS app and local protocol gateway.

SEO, AEO and GEO audit MCP server for Claude Code, Cursor and Codex. Audit a page as AI search crawlers see it, get a fix pack, apply and verify. Free, no API key.

Hourtick for herdr: connect your herdr agents to Hourtick, start one on a task, and track and log the time they work.

Multi-agent GitHub PR review bot — security, performance & test agents with a measured eval pipeline

Personal agent that hunts remote design jobs, ranks them against hard rules, and drafts LinkedIn outreach with human approval.

Typed JavaScript/TypeScript SDK for the WebChatAgent chat API: streaming, file upload, feedback, lead forms and live chat. Zero dependencies.

Fully automatic proxy upkeep: finds fresh server lists, tests, drops dead ones, switches with no human input. Windows (.exe) + Linux (.deb/.rpm/Arch/AppImage/tarball). Vibe-coded.

AI estimating assistant for LV switchgear: computer vision, LLM spec extraction, calibrated confidence, human-in-the-loop review