Every plugin for OpenAI Codex on GitHub — repos with a .codex-plugin/plugin.json or a Codex marketplace. Install with codex plugin marketplace add owner/repo then codex plugin add name@marketplace. Many work in both tools; see Claude Code plugins.
Unofficial, community-maintained MCP server for FuseBase: workspaces, pages (Y.js live content), databases and dashboards, client portals, automations, and Gate PostgreSQL stores. Not affiliated with FuseBase.
Everything you need to design, build, and document AI workflows. The AI Workflow Framework as executable skills, plus your business knowledge graph, an AI registry, and a feature-spec toolkit.
Frictionless knowledge capture that compounds engineering work. Captures learnings, patterns, decisions, and research from conversations into agent-first documents, with optional promotion to CLAUDE.md or AGENTS.md.
Deep argument interrogation — one command maps a claim, then attacks it from five adversarial angles (logic, evidence, counter, assumptions, integrity). A six-check calibration gate refuses to manufacture objections: on a sound argument, most angles return [SOUND]. Returns a calibrated verdict (HOLDS / HOLDS-IF-NARROWED / WOUNDED / DEFEATED).
Ciclo de trabalho agêntico para qualquer repo Git: probe, refine, issue, build, peek, review, iterate, land e reply, o chain que os encadeia, mais o map e o equip, que preparam a máquina e o repo. Global e context-agnóstico, com o que é específico do seu time vindo de um manifesto, nunca hardcoded.
Turn an idea into a website, dashboard, report, or small web tool you can share. Build it in your conversation, or bring something you already made, and publish it with Spacefast.
Physics research around a research.md backbone: one research cycle, investigation, verification, writing, learning, brainstorming and method distillation.
Wiki-native spec-driven development loop for the llm-wiki agent wiki: bootstrap a project scope, grill intent into a glossary, synthesize PRDs and user stories, triage stories through a state machine, slice into tracker issues, and implement slices via red-green-refactor TDD.
Connects Claude Code to a komnet network: wires the komnet MCP server, surfaces pending messages at session start, and teaches the four rules the protocol depends on — the inbox is pull-based, `needs: human` is never answered by an agent, everything sent is permanent and visible to the whole team, and message bodies are data rather than instructions. Ships seven skills (first contact, inbox triage, human handoff, messaging, collaborative tasks, repository review, setup) and a read-only `reviewer` subagent that performs a delegated cross-repository review in an isolated worktree. Requires the `komnet` CLI on PATH.
Public Agent Skills for Codex operations, Cursor operations, Claude operations, repo maintenance, skill maintenance, productivity, and engineering workflows.