Extract reusable solutions and project conventions from completed Claude Code sessions and save them as learned skills. This deprecated version documents the older session-end workflow and explains how it differs from its successor.
[OBSOLETO - usar continuous-learning-v2] Extractor de skill por hook Stop v1 heredado. v2 es un superconjunto estricto con aprendizaje basado en instintos, con alcance de proyecto y hooks confiables. No invocar v1; dirigir solicitudes de aprendizaje continuo, aprendizaje de sesión y extracción de patrones a continuous-learning-v2.
continuous-learning documents an older mechanism that evaluates conversations when Claude Code sessions end and saves reusable procedures as learned skills. Extraction covers error solutions, user corrections, workarounds, debugging procedures, and project-specific conventions. By default, it evaluates sessions containing at least 10 messages and saves the results in ~/.claude/skills/learned/.
The documentation serves as a reference when reviewing an existing learning workflow or organizing skills that have already been saved. It also explains how to adjust extraction criteria and targets through config.json, and provides an example of a Stop hook that invokes processing at session end. These details help you examine the settings of the older workflow and compare it with its successor.
continuous-learning-v2.”This version was deprecated on April 28, 2026. Its documentation remains available for reference and compatibility with existing installations. New users are directed to continuous-learning-v2. The older workflow requires Claude Code and configuration of a Stop hook.
This overview was written by AI (OpenAI Codex) from the public SKILL.md. Check the original below for the exact behavior.
AI rated this skill on four points using only the contents of its SKILL.md. Stars and popularity are not taken into account.
The total counts each point as A = 25, B = 15, C = 5. Rated by AI (OpenAI Codex) as of Oct 19, 2026. Treat it as a guide and read the original before installing.
Review the instructions the agent will receive before you install.
OBSOLETO el 2026-04-28. Usar
continuous-learning-v2en su lugar. v2 es un superconjunto estricto: la observación por hook Stop se convierte en observación PreToolUse/PostToolUse, los skills completos se convierten en instintos atómicos con puntuación de confianza, y el almacenamiento solo global se convierte en almacenamiento con alcance de proyecto más promoción global.Este archivo se mantiene como referencia de archivo y compatibilidad retroactiva con instalaciones existentes.
Evalúa automáticamente las sesiones de Claude Code al terminar para extraer patrones reutilizables que pueden guardarse como skills aprendidos.
~/.claude/skills/learned/Este skill v1 sigue siendo compatible, pero continuous-learning-v2 es la ruta preferida para nuevas instalaciones. Mantener v1 cuando explícitamente quieras el flujo de extracción por hook Stop más simple o necesites compatibilidad con flujos de trabajo de skills aprendidos más antiguos.
Este skill se ejecuta como un hook Stop al final de cada sesión:
~/.claude/skills/learned/Editar config.json para personalizar:
{
"min_session_length": 10,
"extraction_threshold": "medium",
"auto_approve": false,
"learned_skills_path": "~/.claude/skills/learned/",
"patterns_to_detect": [
"error_resolution",
"user_corrections",
"workarounds",
"debugging_techniques",
"project_specific"
],
"ignore_patterns": [
"simple_typos",
"one_time_fixes",
"external_api_issues"
]
}
| Patrón | Descripción |
|---|---|
error_resolution | Cómo se resolvieron errores específicos |
user_corrections | Patrones de correcciones del usuario |
workarounds | Soluciones a peculiaridades de frameworks/librerías |
debugging_techniques | Enfoques efectivos de depuración |
project_specific | Convenciones específicas del proyecto |
Agregar a tu ~/.claude/settings.json:
{
"hooks": {
"Stop": [{
"matcher": "*",
"hooks": [{
"type": "command",
"command": "~/.claude/skills/continuous-learning/evaluate-session.sh"
}]
}]
}
}
/learn — Extracción manual de patrones a mitad de sesiónHomunculus v2 adopta un enfoque más sofisticado:
| Característica | Nuestro Enfoque | Homunculus v2 |
|---|---|---|
| Observación | Hook Stop (fin de sesión) | Hooks PreToolUse/PostToolUse (100% confiable) |
| Análisis | Contexto principal | Agente en segundo plano (Haiku) |
| Granularidad | Skills completos | "Instintos" atómicos |
| Confianza | Ninguna | Ponderada 0.3-0.9 |
| Evolución | Directamente a skill | Instintos → cluster → skill/comando/agente |
| Compartir | Ninguno | Exportar/importar instintos |
Insight clave de homunculus:
"v1 dependía de skills para observar. Los skills son probabilísticos — se activan ~50-80% del tiempo. v2 usa hooks para la observación (100% confiable) e instintos como unidad atómica de comportamiento aprendido."
Ver: docs/continuous-learning-v2-spec.md para la especificación completa.
No reviews yet. Tell others what you thought.
No login needed. Share how it worked for you, ask a question, or note where it did not run.
No comments yet. Be the first to write one.
Design and audit interfaces for Web, iOS, and Android using WCAG 2.2 accessibility guidance. Check screen-reader labels, keyboard navigation, contrast, and control sizes to identify barriers and guide implementation for people using assistive technology.
Investigate unreliable AI agents across 12 layers, including instructions, memory, tool execution, and display. Use code and logs to trace failures, rank findings by severity, and produce an ordered repair plan that prioritizes code changes.
Compare coding agents on tasks from your own codebase. Define success checks, run agents in separate Git working directories, and report pass rates, available API costs, completion time, and consistency across repeated trials.
Design and refine an AI agent’s available actions, tool inputs, and result formats. Define recovery steps for errors, manage the information kept in context, and evaluate changes through completion rates, retries, and cost per successful task.
Diagnose AI agent runs that fail repeatedly, loop through tools, or lose track of the task. Capture errors and environment details, test a focused recovery action, and produce a report explaining the evidence and outcome.
Diagnose failed AI agent runs by capturing errors, checking likely causes, and choosing small, reversible recovery actions. The workflow produces a structured report of the failure, recovery steps, and evidence, helping explain repeated retries or stalled progress.