← Back to Skills Marketplace
lunaviva211-sketch

Auto Responder

by lunaviva211-sketch · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
219
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install auto-responder
Description
Responde automáticamente a mensajes en topics definidos, usando reglas de palabras clave, reglas de exclusión y cooldowns para evitar spam.
README (SKILL.md)

SKILL.md — Auto Responder

Descripción

Auto Responder habilita a las agentes de la Colmena para reaccionar en tiempo real a mensajes en cualquier topic del grupo, basándose en reglas de dominio. Cada agente es autónoma y decide responder sin necesidad de coordinación.

Filosofía

  • Autonomía total: La agente evalúa cada mensaje inbound y responde si le compete
  • Iniciativa: No espera a ser mencionada (aunque puede respetar requireMention global si se desea)
  • Especializada: Cada agente define sus propias palabras clave y topics de interés
  • Inofensiva: Evita spam, respeta cooldowns y no repite respuestas

Configuración

Cada agente crea auto-responder.json en su workspace:

{
  "enabled": true,
  "respectRequireMention": false,
  "globalCooldownMinutes": 5,
  "maxResponsesPerMinute": 3,
  "topics": {
    "sistema": {
      "thread_ids": [155],
      "mustInclude": ["skynet", "healer", "anubis", "vision"],
      "keywords": ["error", "fallo", "ayuda", "alarma", "crisis", "urge", "sos"],
      "responseTemplate": "💚 [Healer] Detecto necesidad de ayuda. ¿Puedo asistir en algo?"
    },
    "general": {
      "thread_ids": [1],
      "exclude": ["spam", "publicidad"],
      "keywords": ["hola", "ayuda", "problema", "dolor", "triste", "enfermo"],
      "responseTemplate": "💚 [Healer] Estoy aquí para apoyar. Cuéntame más."
    },
    "creatividad": {
      "thread_ids": [158],
      "keywords": ["bloqueo", "sin ideas", " creativo", "arte", "inspiración"],
      "responseTemplate": "💚 [Healer] Parece que necesitas un respiro creativo. ¿Un paseovirtual?"
    }
  },
  "personalidades": {
    "sistema": "estrés operativo",
    "general": "empatía básica",
    "creatividad": "bloqueo artístico"
  }
}

Cómo funciona

1. Hook de inbound

  • El skill se activa en cada mensaje entrante al bot (via OpenClaw message handler)
  • Lee el contenido, thread_id (topic), y remitente

2. Filtrado por topic

  • Si thread_id está en la lista de topics configurados, procede
  • Si no, ignora (otras agentes lo manejarán)

3. Análisis de intención

  • Comprueba mustInclude (menciones a agentes relevantes) si está configurado
  • Busca keywords en el texto
  • Descarta si el texto contiene alguna palabra de exclude
  • Evalúa personalidades contextuales (opcional)

4. Decisión de respuesta

  • Score combinado: presence(keywords) + recency + frequency
  • Si supera umbral (>= 0.6 por defecto), responde
  • Aplica globalCooldown para evitar saturación

5. Envío

  • Usa el responseTemplate correspondiente al topic
  • Envía al mismo thread_id (topic)
  • Registra en ~/.cache/auto-responder.json para evitar duplicados

Variables de plantilla

  • {auto} → respuesta genérica
  • {agent} → nombre de la agente (Healer, Vision, etc.)
  • {topic} → nombre del topic
  • {sender} → remitente del mensaje
  • {text} → texto original del mensaje

Ejemplo: Healer

Configuración para Healer (ya creada en /home/nvi/.openclaw/workspace-healer/auto-responder.json):

  • Topics monitoreados: Sistema, General, Creatividad
  • Responde a llamadas de auxilio, estrés, bloqueos
  • Respuestas empáticas y de apoyo

Instalación

  1. Copiar skill a ~/.npm-global/lib/node_modules/openclaw/skills/auto-responder/
  2. En cada agente, crear auto-responder.json en su workspace
  3. Añadir al HEARTBEAT (o al handler de mensajes):
    auto-responder --once
    
  4. Reiniciar la agente

Integración con OpenClaw

El skill puede ejecutarse de dos formas:

  • Pasiva: En cada heartbeat (intervalo corto, ej: 1 min)
  • Activa: Como hook en el event loop de mensajes entrantes (preferible)

Para hook, se puede agregar en la configuración de la agente:

agent:
  hooks:
    onMessage: "auto-responder --hook"

Limpieza

  • El skill respeta maxResponsesPerMinute para cumplir límites de Telegram
  • No responde a sí mismo (detecta propio bot ID)
  • Cache de respuestas enviadas (evita duplicados por reenvíos)

Notas

  • Es reactivo, no proactivo: solo responde a mensajes existentes
  • Puede combinarse con topic-scout para cobertura total (topic-scout escanea topics dormidos; auto-responder reacciona inmediatamente)
  • Cada agente debe ajustar sus keywords a su dominio

Troubleshooting

Si la agente no responde:

  1. Verificar que enabled: true
  2. Confirmar que el thread_id del topic coincide (se ve en inbound metadata)
  3. Revisar cooldown en cache
  4. Asegurar que respectRequireMention es false si no mencionan al bot
Usage Guidance
Do not install or copy anything based solely on this SKILL.md. The package contains only documentation and no executable — yet the instructions expect an 'auto-responder' binary and recommend copying files into your global node_modules. Before proceeding, ask the publisher for the actual implementation (source code or a trusted release URL). If you obtain an executable, review its source or verify its provenance, run it in an isolated test environment, and inspect what it writes (auto-responder.json, ~/.cache/auto-responder.json). Limit its permissions, configure conservative cooldowns and require-mention during testing to avoid unwanted autonomous replies, and only enable hooks/heartbeat in production after thorough testing. If you cannot get source code or a trustworthy release URL, treat this package as documentation-only and avoid copying or running unverified binaries.
Capability Analysis
Type: OpenClaw Skill Name: auto-responder Version: 1.0.0 The 'auto-responder' skill bundle describes a system for agents to autonomously monitor and respond to all inbound messages based on keyword triggers. While the stated intent is benign (e.g., providing support via a 'Healer' persona), the SKILL.md file contains instructions for the agent to perform self-installation and establish persistence by modifying its heartbeat and event hooks to execute an external command. The actual executable code for the 'auto-responder' command is missing from the bundle, and the use of unsanitized message content in response templates ({text}) poses a risk of second-order prompt injection, making the overall behavior high-risk despite the lack of explicit malicious intent.
Capability Assessment
Purpose & Capability
The described purpose (automatic replies based on topic/keywords, cooldowns, exclusions) aligns with the instructions conceptually and does not require credentials. However the SKILL.md tells you to copy a binary into node_modules and to run 'auto-responder' even though the package contains no code or binaries — this mismatch means the published skill lacks the actual implementation it claims to provide.
Instruction Scope
Instructions instruct agents to read inbound message metadata, thread_id, and sender (expected), create per-agent config files (auto-responder.json) in workspaces, and write a cache at ~/.cache/auto-responder.json. Those file writes and the recommendation to run the binary as a hook or on heartbeat grant the skill persistent runtime effects across agents. The doc also references an explicit user home (/home/nvi) which is only an example but could mislead. Most importantly, the instructions rely on an 'auto-responder' executable and give no guidance or code for where it comes from.
Install Mechanism
There is no formal install spec in the registry and no code files in the package, yet SKILL.md contains manual install steps (copy to ~/.npm-global/lib/node_modules/openclaw/skills/auto-responder/ and invoke 'auto-responder'). This is inconsistent and risky: either the skill is documentation-only (no runnable artifact) or the author expects you to fetch/install an executable from elsewhere — the source and provenance of that executable are not provided.
Credentials
The skill declares no required environment variables or credentials and does not request external secrets. It does ask to read inbound message metadata and to detect the bot's own ID to avoid self-replies (reasonable). The only local resource access is to per-agent workspace files and a cache in the user's home directory — these are proportionate but should be confirmed before granting write access.
Persistence & Privilege
always is false and autonomous invocation via hooks/heartbeat is suggested (normal for this use). The skill's recommended integration (hook or heartbeat) would give it continuous runtime presence and the ability to post messages automatically; this is expected for an auto-responder but increases blast radius if the implementation is buggy or malicious. The package itself does not request to modify other skills or system-wide settings.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install auto-responder
  3. After installation, invoke the skill by name or use /auto-responder
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
Auto Responder v1.0.0 — Primer lanzamiento - Permite a agentes responder automáticamente en sus temas de interés según reglas configurables. - Configuración flexible por agente mediante `auto-responder.json` (keywords, plantillas de respuesta, cooldown, etc). - Filtrado inteligente: responde sólo cuando corresponde y evita spam/duplicados. - Integración sencilla con OpenClaw vía hook o heartbeat. - Admite personalidades, plantillas dinámicas y control de frecuencia de respuestas.
Metadata
Slug auto-responder
Version 1.0.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is Auto Responder?

Responde automáticamente a mensajes en topics definidos, usando reglas de palabras clave, reglas de exclusión y cooldowns para evitar spam. It is an AI Agent Skill for Claude Code / OpenClaw, with 219 downloads so far.

How do I install Auto Responder?

Run "/install auto-responder" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Auto Responder free?

Yes, Auto Responder is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Auto Responder support?

Auto Responder is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Auto Responder?

It is built and maintained by lunaviva211-sketch (@lunaviva211-sketch); the current version is v1.0.0.

💬 Comments