continuous-learning-v2

Captures Claude Code session activity via hooks and evolves observed patterns into reusable instincts, skills, and commands.

Updated Mar 18, 2026
One-click install
npx skills add https://github.com/freedom909/real-estate-saas --skill continuous-learning-v2-freedom909
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: continuous-learning-v2
Source: https://github.com/freedom909/real-estate-saas/tree/main/.trae/skills/continuous-learning-v2
Command: npx skills add https://github.com/freedom909/real-estate-saas --skill continuous-learning-v2-freedom909

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve? Valuable patterns from your Claude Code sessions—user corrections, error resolutions, repeated workflows—are lost when sessions end. This Skill observes tool usage deterministically through PreToolUse/PostToolUse hooks, distills recurring behaviors into atomic confidence-scored "instincts", and clusters them into reusable skills, commands, and agents, with project-scoped storage preventing cross-project contamination. ## Core Features & Use Cases - Hook-Based Observation: Captures 100% of tool calls via PreToolUse/PostToolUse hooks into project-scoped observations.jsonl files, with secret scrubbing and automated-session filtering. - Instinct Management CLI: Provides status, import, export, evolve, promote, projects, and prune commands for managing confidence-weighted instincts across project and global scopes. - Background Observer Agent: A Haiku-based observer loop analyzes accumulated observations on an interval or SIGUSR1 signal, guarded by active-hours, cooldown, and idle-detection gates. - Use Case: After repeatedly correcting Claude to prefer functional React components, the system creates a project-scoped instinct with rising confidence; when the same pattern appears in multiple projects at high confidence, it is promoted to a global instinct and evolved into a reusable skill. ## Quick Start Enable the observation hooks in your Claude Code settings and ask the assistant to run the instinct-status command to review learned patterns for the current project.

Frequently Asked Questions about continuous-learning-v2

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I set up automatic learning from Claude Code sessions?▼

Register the observe.sh script as PreToolUse and PostToolUse hooks in your Claude Code settings.json, or install it as a plugin so hooks.json loads automatically. Observations then accumulate per project and are analyzed by the background observer.

How do I review and manage learned instincts?▼

Run the instinct-cli.py status command or the /instinct-status command to list project-scoped and global instincts grouped by domain with confidence bars. You can then export, import, promote, or prune instincts from the same CLI.

Does the learning system work across multiple projects?▼

Yes, version 2.1 scopes instincts per project using a hash of the git remote URL, so React patterns stay in your React project. Universal patterns can be promoted to global scope when seen in two or more projects with confidence of 0.8 or higher.

Why are hooks used instead of skills for observation?▼

Hooks fire deterministically on every tool call, while skills activate probabilistically only 50-80% of the time based on model judgment. This guarantees no session patterns are missed during observation.

What are the limitations of the background observer?▼

The observer is disabled by default and requires at least 20 observations before analysis runs. On Windows it skips Claude analysis due to a known non-interactive hang unless ECC_OBSERVER_ALLOW_WINDOWS is set, and cycles are gated by active hours, cooldowns, and idle detection.