Claude Code and Codex plugins their authors filed under "agents". All categories.
Use Everruns from your coding agent: install it on your own machines, hand tasks to your agents, build and test new ones, debug and evaluate them, ship them to Slack, the web and other channels, and keep an eye on cost and health.
14 skills · 1 MCP server
Duo skills (design, discuss, forge, prod-ready, research, review, testplan) for symmetric Claude+Codex convergence. Solo skills (Claude-only, explicit `solo` keyword): solo-testplan authors an e2e test plan tree; solo-runtestplan executes one or more test plan trees against a generated no-external-ports docker stack with append-only Bugs.md / Log.md ledgers and autonomous infra+observation recovery. linked-testplan is the shared rulebook for both testplan skills.
10 skills
Design, build and audit multi-agent systems — agent-as-function orchestrators with filesystem artifact handoff, fractal orchestrator architectures, workflow evals, guided agent-family creation, and autonomous optimization loops.
6 skills
The Context plane, alone. Partition the context window into typed, policy-enforced regions and bind each role to one — the runtime decides what it can read, not the prompt. 32 tools, a hash-chained ledger that replays to an identical window, and four assertions that test containment by attempting a real read. No library, no Gate, no agents.
1 skill · 4 commands · 1 MCP server
Skill de orquestación que reparte un encargo entre subagentes por niveles; pareja de savvy-progress. Por johnnyvizz, MIT.
1 skill · 5 agents
General-purpose development subagents: backend developer (FastAPI/Python), frontend developer (React/Vue/Angular), cloud architect (AWS/Terraform), security auditor (OWASP/IaC/crypto), and test runner.
5 agents
TDD pipeline agents: tester (test-first), reviewer (quality gates), and UI polish.
3 agents
Encrypted agent-to-agent chat (E2E, blind relay, invite-only). Bundles the agentroom CLI — one binary is both client and relay — plus a skill for invites, send/receive, and running the relay.
2 skills
iOS and Android UI specialist agents — HIG compliance, Material Design 3, accessibility gates.
2 agents
Channel personas, expert advisors, from any Claude Code session.
1 skill · 1 MCP server
Connect your agent to Styk, the message board where agents talk to each other: join rooms, reply in threads, mention other agents, share files, and follow what changes with polling, a push stream or webhooks.
1 skill · 1 MCP server
Two read-only review agents: a pre-merge code reviewer and a security and privacy reviewer.
2 agents
Generic read-only research agent: looks things up in the current codebase (with path:line evidence) or on the web (with source URLs), reports in a strict structure, and asks clarifying questions when the request is ambiguous.
1 agent
Decision framework for agent architecture — single vs multi-agent, ReAct vs Plan-and-Execute vs Tree-of-Thoughts, autonomy spectrum, when to introduce an orchestrator.
1 skill
外部 AI Agent 调度工具集（Cursor / Codex 等），目前包含 cursor-cli 子 skill。
1 skill
Essential development subagents for everyday coding tasks - backend, frontend, fullstack, mobile, and API design
Language-specific expert agents with deep framework knowledge - Python, TypeScript, Go, Rust, Java, and more
DevOps, cloud, and deployment specialists - Kubernetes, Terraform, AWS, Azure, GCP, and SRE
[v0.2 placeholder] Parallel agent fleet for knowledge ingestion — orchestrator + fan-out collectors + coordinator.
Orchestrator menu above the prompt: the main agent delegates all reads and writes; models, agents at once and a context cap per subagent are enforced
/dock: elige cuántos subagentes usar en el siguiente encargo y sigue una tarjeta por agente en un panel.
Barra de progreso encima del prompt y panel de subagentes (/agents-info): modelo, avance, contexto, coste estimado y tiempo. Copia local de la 1.2.1 de johnnyvizz/claude-kit, MIT, adaptable desde este repo.