continuous-learning

Extract reusable patterns from Claude Code sessions and save them as learned skills.

Updated Aug 27, 2026
One-click install
npx skills add https://github.com/overthestream/agent-dotfile --skill continuous-learning-overthestream
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: continuous-learning
Source: https://github.com/overthestream/agent-dotfile/tree/main/claude/skills/continuous-learning
Command: npx skills add https://github.com/overthestream/agent-dotfile --skill continuous-learning-overthestream

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires jq.

What problem does it solve?

Automatically extracts reusable patterns from Claude Code sessions and saves them as learned skills, reducing manual effort and enabling faster improvement across projects.

Core Features & Use Cases

  • Stop-hook evaluation at the end of each Claude Code session to analyze transcripts.
  • Pattern detection across sessions to identify reusable tactics (error_resolution, user_corrections, workarounds, debugging_techniques, project_specific).
  • Persists discovered patterns as new learned skills under ~/.claude/skills/learned/ for quick reuse in future sessions.
  • Configurable via config.json to tailor extraction thresholds and destinations.

Quick Start

Enable the Stop hook and run Claude Code sessions to automatically extract and save patterns to your learned-skills directory.

Frequently Asked Questions about continuous-learning

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

FAQPage Schema
How do I automatically extract reusable patterns from Claude Code sessions?

You can extract patterns from Claude Code sessions by using a session-end hook that analyzes transcripts and saves identified tactics as learned skills under ~/.claude/skills/learned/. This automates pattern extraction without manual effort.

How does session-end pattern extraction work for Claude Code transcripts?

Session-end pattern extraction works by applying a stop hook that evaluates transcripts after a Claude Code session concludes. It detects patterns like error_resolution and debugging_techniques, then persists them for future reuse.

Can I configure the extraction threshold for Claude Code session patterns?

Yes, you can configure the extraction threshold and other settings via a config.json file. This file lets you define min_session_length, auto_approve, learned_skills_path, and the specific patterns_to_detect.

Do I need jq installed to extract learned skills from Claude Code sessions?

Yes, jq is required as a dependency to parse and extract learned skills from Claude Code session transcripts. The evaluate-session.sh script relies on it to process the JSON data.

What types of patterns can I detect from Claude Code session transcripts?

You can detect reusable patterns such as error_resolution, user_corrections, workarounds, debugging_techniques, and project_specific patterns from your Claude Code session transcripts.

How do I stop Claude Code from extracting patterns on sessions that are too short?

To prevent pattern extraction on short sessions, set the min_session_length parameter in your config.json file. The evaluate-session.sh script checks this threshold before analyzing the transcript.