ptreezh/stigmergy-CLI-Multi-Agents

Última actividad de origen registrada: 12 de octubre de 2026 a las 00:58

Catálogo de SkillsMP actualizado: 12 de noviembre de 2026 a las 12:52

skills recopiladas: 35

Estrellas en GitHub: 2
Forks en GitHub: 0

Instalar con un asistente de IA

Copia este prompt en el asistente de IA que estés usando.

Follow https://skillsmp.com/skill-install/prompt.md to install Agent Skills from https://github.com/ptreezh/stigmergy-CLI-Multi-Agents.

README del repositorio

Archivo de origen del README: README.md

Leer el README original

Mostrar README completo

# Stigmergy CLI ![Stigmergy CLI — write a skill once, run it in every AI CLI](./assets/banner.png) > **Write a skill once. Run it in every AI CLI.** > Stigmergy is a portability + coordination layer across the AI CLIs you already use. Install an agent skill in one place, and Claude, Gemini, Qwen, iFlow, Qoder, CodeBuddy, Codex, KiloCode, and opencode can all use it — and share session context instead of making you re-explain everything. [![npm version](https://badge.fury.io/js/stigmergy.svg)](https://www.npmjs.com/package/stigmergy) [![CI](https://github.com/ptreezh/stigmergy-CLI-Multi-Agents/actions/workflows/ci.yml/badge.svg)](https://github.com/ptreezh/stigmergy-CLI-Multi-Agents/actions/workflows/ci.yml) [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT) [![npm downloads](https://img.shields.io/npm/dm/stigmergy)](https://www.npmjs.com/package/stigmergy) **English** · A portability + coordination layer for multiple AI CLIs: install skills once and use them anywhere, recover session context across CLIs, and (optionally) route or fan out one task to several agents. **中文** · 多 AI CLI 的可移植与协作层：技能装一次、处处可用；会话上下文跨 CLI 恢复；可选地把一个任务路由或并发派发给多个 agent。 ## The problem You already run several AI CLIs (Claude, Gemini, Qwen, iFlow, Qoder, CodeBuddy, Codex, KiloCode, opencode…). Each keeps its **own** skills directory (`~/.claude/skills`, `~/.qwen/skills`, …), its **own** session history, and its **own** conventions. So you reinstall the same skill everywhere, re-explain the same context every time you switch tools, and cannot hand a task from one CLI to another. Stigmergy centralizes that: **one skill store, one session bridge, one optional router.** ## 60-second demo (real output, trimmed) ```console $ stigmergy --version 1.11.0 # Detect what is ACTUALLY installed (probes PATH; reports only what it finds): $ stigmergy scan Found 9 CLI tools: Claude CLI Version: 2.1.295 (Claude Code) Gemini CLI Version: 0.63.0 Qwen CLI Version: 0.25.0 iFlow CLI Version: 0.5.19 OpenCode AI CLI Version: 1.18.35 Qoder CLI Version: 1.1.66 CodeBuddy CLI Version: 2.162.0 KiloCode CLI Version: 7.8.8 OpenAI Codex CLI Version: codex-cli 0.146.0 (copilot, kode: probed, NOT found — nothing above is hard-coded) # Every skill Stigmergy has unified, across all your CLIs: $ stigmergy skill list Installed skills (352): [UNIVERSAL] universal: algorithmic-art Creating algorithmic art using p5.js with seeded randomness… brainstorming You MUST use this before any creative work… … ``` The `scan`/`status` lists are produced by probing your `PATH`; `skill list` counts only skills actually installed. **Letter-for-letter transcript:** [`examples/quickstart.md`](./examples/quickstart.md). ## The one thing to try ```bash # Install a skill from any public GitHub repo — once: stigmergy skill install anthropics/skills # It is now readable by every CLI that has Stigmergy hooks deployed: stigmergy skill read pdf # prints the SKILL.md for the calling agent ``` Cross-CLI skill and session portability is the core, demand-backed use case (see the comparison below): sibling projects prove the niche is real, yet multi-CLI skill sharing remained largely un-served. ## How it works ![Stigmergy architecture — a skill store, a shared session bus, and an optional router spanning every supported CLI](./assets/architecture.png) Skills land in one **shared store** that every CLI reads through its deployed hooks; the **session bus** lets one CLI pick up where another left off; the optional **router** fans a single task out to several agents. No CLI is replaced — they are coordinated. ## Install Requires Node.js >= 16. ```bash npm install -g stigmergy@beta # current npm channel stigmergy setup # scan CLIs, deploy hooks, install shared skills ``` > **Status — read before production use.** Stigmergy is effectively **pre-1.0**: the npm `latest` tag currently points at a `1.10.10-beta.5` build, while this repository is at `1.11.0`. A true stable release (cutting `1.11.0` and moving betas to the `@beta` tag) is on the roadmap. Recent work hardened cross-CLI skill sync so it writes only the memory filenames each CLI actually reads — see [CHANGELOG.md](./CHANGELOG.md). Pin `@beta` and check the changelog before relying on it. ## How it compares | Project | Focus | Cross-CLI skill portability | Cross-CLI session recovery | |---|---|---|---| | **Stigmergy** | Portable skills + session bridge across many AI CLIs | ✅ install once, run anywhere | ✅ shared session bus | | `rtk-ai/icm` (543★) | Cross-agent memory | partial (memory, not skills) | partial | | `Kaseban/baton` (26★) | Session conversion | no | yes (converts one session) | | `opencode` | One AI coding agent for the terminal | no (single tool) | no | | `Cline` (69,556★) | Autonomous coding agent in the IDE | no (single tool) | no | Star counts measured 2026-10-10. Single-tool projects are excellent — they just do not span multiple CLIs by design. Stigmergy **coordinates** the tools you already use rather than replacing them. ## Other features (Experimental) These ship today but are newer and less proven than the core portability path. Treat them as experimental: - **Smart routing** — `stigmergy call "<task>"` chooses a CLI for the prompt. - **Concurrent execution** — `stigmergy concurrent "<task>"` fans one task out to several agents. - **Project Status Board** — `stigmergy interactive` keeps a per-directory board at `.stigmergy/status/PROJECT_STATUS.md`. - **Agent observatory** — `stigmergy wiki-scan` reads agent memory files and summarizes real activity (evidence-first: reports only detected agents). - **Gateway** — `stigmergy gateway --feishu | --telegram | --slack` exposes CLIs over chat platforms. - **Soul self-evolution** — `stigmergy soul evolve | reflect | co-evolve | compete`. ## Supported CLIs Claude · Gemini · Qwen · iFlow · Qoder · CodeBuddy · Codex · KiloCode · opencode · WorkBuddy Desktop. Add or refresh adapters with `stigmergy install <tool>`. ## Documentation - [Project Constitution](./docs/project-constitution.md) — what this project is and its principles - [Tutorials: multi-agent workflows](./docs/tutorials/multi-agent-workflows.md) - [AI products ontology](./docs/ai-products-ontology.md) — supported agents and their configs - [Self-reporting architecture](./docs/self-reporting-architecture.md) - [Launch post (draft, unpublished)](./docs/archive/launch-2026-10-07.md) - [Changelog](./CHANGELOG.md) · [Contributing](./CONTRIBUTING.md) · [Security](./SECURITY.md) ## Roadmap - Cut a true **stable** release and move betas to the `@beta` tag. - Curate `docs/` (in progress) and keep the surface small. ## Demo A visual asciinema/GIF demo is on the roadmap (deferred from 1.11.0 — the maintainer's environment was missing asciinema at release time, and the project does not ship Windows binaries). The verified text transcript in [`examples/quickstart.md`](./examples/quickstart.md) is the canonical proof of behaviour; the visual demo will be added in a follow-up release once the tooling is available. ## Contributing See [CONTRIBUTING.md](./CONTRIBUTING.md). The project is maintained by a human maintainer working with AI collaborators; contributions are welcome. ```bash git clone https://github.com/ptreezh/stigmergy-CLI-Multi-Agents.git npm install && npm run build:orchestration && npm test ``` ## License MIT — see [LICENSE](./LICENSE) (Chinese translation: [LICENSE.zh.md](./LICENSE.zh.md)). --- _Version: 1.11.0 (development) · Latest npm: 1.10.10-beta.5_

Cargando el README del repositorio…

Skills en este repositorio

clasificación pendiente

skill

ocupación

descripción

actualizado

soul-co-evolve

sin clasificar

协同进化 - 基于双Agent循环的多CLI协作学习

Idioma del texto original: chino

2026-10-12

soul-compete

sin clasificar

竞争进化 - 基于双Agent循环的方案对比与最优选择

Idioma del texto original: chino

2026-10-12

soul-evolution

sin clasificar

自主进化 - 基于双Agent循环的自我学习与进化

Idioma del texto original: chino

2026-10-12

soul-reflection

sin clasificar

自我反思 - 基于双Agent循环的分析与学习提取

Idioma del texto original: chino

2026-10-12

soul-co-evolve

sin clasificar

协同进化 - 基于双Agent循环的多CLI协作学习

Idioma del texto original: chino

2026-10-12

soul-compete

sin clasificar

竞争进化 - 基于双Agent循环的方案对比与最优选择

Idioma del texto original: chino

2026-10-12

soul-evolution

sin clasificar

自主进化 - 基于双Agent循环的自我学习与进化

Idioma del texto original: chino

2026-10-12

soul-reflection

sin clasificar

自我反思 - 基于双Agent循环的分析与学习提取

Idioma del texto original: chino

2026-10-12

soul-co-evolve

sin clasificar

协同进化 - 基于双Agent循环的多CLI协作学习

Idioma del texto original: chino

2026-10-11

soul-compete

sin clasificar

竞争进化 - 基于双Agent循环的方案对比与最优选择

Idioma del texto original: chino

2026-10-11

soul-evolution

sin clasificar

自主进化 - 基于双Agent循环的自我学习与进化

Idioma del texto original: chino

2026-10-11

soul-reflection

sin clasificar

自我反思 - 基于双Agent循环的分析与学习提取

Idioma del texto original: chino

2026-10-11

two-agent-loop

sin clasificar

Two-agent loop - the foundational execution mechanism for all evolution workflows. One agent executes, another reviews and iterates; autonomous evolution, collaborative evolution, and competitive evolution all depend on this pattern.

Idioma del texto original: Varios idiomas

2026-10-11

cli-integration-sop

sin clasificar

CLI integration Standard Operating Procedure - a standardized, step-by-step checklist for integrating a new CLI tool into the Stigmergy Multi-Agents CLI Collaboration System. Use when adding adapter support for a new AI CLI tool.

Idioma del texto original: inglés

2026-10-11

complex-task-decomposition-with-system-engineering

sin clasificar

Complex task decomposition with systems-engineering thinking - handle complex multi-step tasks via persistent file management, task decomposition, and concurrent agents for high-efficiency execution.

Idioma del texto original: chino

2026-10-11

planning-with-files

sin clasificar

Implements Manus-style file-based planning for complex tasks. Creates task_plan.md, findings.md, and progress.md. Use when starting complex multi-step tasks, research projects, or any task requiring >5 tool calls.

Idioma del texto original: inglés

2026-10-11

resumesession

sin clasificar

Cross-CLI session recovery and history management skill

Idioma del texto original: inglés

2026-10-11

strict-test-skill

sin clasificar

Strict test skill - verifies the real activation mechanism of the CLI skill system by confirming whether the skill is actually loaded and executed.

Idioma del texto original: Varios idiomas

2026-10-11

verification-first

sin clasificar

Verification-first skill - prevent "simulated instead of actually executed" cognitive errors by verifying before claiming completion, fixed, or passed. Use when about to claim work is done, before committing or creating PRs, or when making assumptions about…

Idioma del texto original: inglés

2026-10-11

cli-integration-sop

sin clasificar

CLI 集成标准操作程序 - 用于将新 CLI 工具集成到 Stigmergy 系统

Idioma del texto original: inglés

2026-03-12

skill-from-masters

sin clasificar

大师技能 - 提供专家研究方法论用于高级分析任务

Idioma del texto original: inglés

2026-03-12

using-superpowers

sin clasificar

强制技能加载协议 - 收到任何消息后必须先检查技能

Idioma del texto original: chino

2026-03-12

two-agent-loop

sin clasificar

双Agent循环 - 所有进化的基础执行机制

Idioma del texto original: chino

2026-03-08

using-superpowers

sin clasificar

Use when starting any conversation - establishes how to find and use skills, requiring skill invocation before ANY response including clarifying questions

Idioma del texto original: inglés

2026-03-08

academic-ant

sin clasificar

学术型行动者网络理论(ANT)研究员技能，融合人格特质与专业方法论，提供严谨的ANT理论分析、本土化案例研究和跨学科对话支持

Idioma del texto original: chino

2026-03-08

soul-auto-compute-hunter

sin clasificar

自主算力狩猎技能 - 自动发现、申请和整合免费LLM API资源

Idioma del texto original: chino

2026-03-08

soul-auto-search-config

sin clasificar

自主搜索引擎配置技能 - 自动检测、申请和配置搜索API Keys

Idioma del texto original: chino

2026-03-08

two-agent-loop

sin clasificar

双Agent循环 - 所有进化的基础执行机制

Idioma del texto original: chino

2026-03-08

soul-auto-evolve

sin clasificar

自主进化技能 - 让本CLI用自己的LLM进行搜索、分析和技能创建

Idioma del texto original: chino

2026-03-06

resumesession

sin clasificar

Cross-CLI session recovery and history management skill

Idioma del texto original: inglés

2026-02-08

academic-illustration

sin clasificar

Generate black-and-white printed illustrations for academic books based on Tufte data-ink ratio principles, Novak concept mapping theory, and causal loop diagrams. Creates diagrams for interdisciplinary theories including core-environment coupling models and…

Idioma del texto original: inglés

2026-02-08

complex-task-decomposition-with-system-engineering

sin clasificar

基于复杂任务分解和系统工程思维的技能，使用持久化文件管理、任务分解和并发智能体实现复杂任务的高效处理

Idioma del texto original: chino

2026-02-08

resumesession

sin clasificar

Cross-CLI session recovery and history management skill

Idioma del texto original: inglés

2026-02-08

strict-test-skill

sin clasificar

严格测试技能 - 用于验证CLI的真实激活机制

Idioma del texto original: chino

2026-01-25

planning-with-files

sin clasificar

Idioma del texto original: inglés

2026-01-21

Mostrando 35 de 35 skills recopiladas.