continuous-learning

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

Updated May 2, 2024
One-click install
npx skills add https://github.com/cman131/EatSomethingSourWhenYoureTired --skill continuous-learning-cman131
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: continuous-learning
Source: https://github.com/cman131/EatSomethingSourWhenYoureTired/tree/main/.claude/skills/continuous-learning
Command: npx skills add https://github.com/cman131/EatSomethingSourWhenYoureTired --skill continuous-learning-cman131

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve? Valuable problem-solving patterns discovered during Claude Code sessions are lost when the session ends, forcing you to rediscover the same fixes and workarounds repeatedly. ## Core Features & Use Cases - Automatic Observation: PreToolUse and PostToolUse hooks capture every tool call as JSONL entries, scoped per project via git remote URL hashing. - Pattern Extraction: Detects error resolutions, user corrections, workarounds, debugging techniques, and project-specific conventions once enough observations accumulate. - Privacy-Aware Storage: Redacts secret-like patterns, truncates sensitive tool payloads, and archives observation files with configurable retention. - Use Case: After several sessions debugging a Node.js API, run the /learn command to extract recurring error-resolution patterns into reusable skill files stored in .claude/skills/learned/. ## Quick Start Ask Claude to set up the continuous-learning observation hooks in your project settings and extract learned patterns from recent sessions.

Frequently Asked Questions about continuous-learning

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

FAQPage Schema
How do I automatically capture patterns from Claude Code sessions?▼

Register the observe.sh script as PreToolUse and PostToolUse hooks in .claude/settings.json. The hooks write tool inputs and outputs as JSONL observations, which the /learn command later analyzes to extract reusable patterns.

How to extract learned skills from Claude Code observations?▼

Run the /learn command after at least 20 observations accumulate (configurable via min_observations_for_extract). Claude reads the observations.jsonl file, detects patterns like error resolutions and workarounds, and saves them as skill files.

Does the observation hook store sensitive data like API keys?▼

The hook redacts secret-like patterns matching passwords, tokens, and API keys before writing observations. Sensitive tools like Bash, Edit, Write, and Read are truncated to 200-character previews, and git remote credentials are stripped.

What happens when the observations file gets too large?▼

When observations.jsonl exceeds the configured max_observation_file_mb (default 10MB), it is automatically moved to an observations.archive directory with a timestamped name. Archived files older than the retention period (default 30 days) are deleted.

Why is the observation hook not recording any data?▼

The hook requires Python (python3 or python) to parse JSON and write observations; without it, observation writing is skipped silently. Also verify the hooks are registered in settings.json and that the script path is correct.