Why does Claude ask for validation before saving a lesson?

Validation is critical to ensure Claude hasn't misidentified the root cause of a problem or focused on a 'red herring' fix that only worked by coincidence.

Can I use Reflect for successful patterns instead of just mistakes?

Yes. While optimized for learning from errors, you can use Reflect to capture any optimal approach or 'best way' to perform a task for future reference.

How do I trigger the Reflect skill?

You can trigger the skill using natural language phrases like 'reflect', 'review our mistakes', 'learn from this', or 'remember this for next time'.

Does this skill work across different projects?

Yes, as long as the skill is active, it will help you build a repository of lessons that Claude can consult whenever it sees a matching activity trigger in your CLAUDE.md.

Where are the lessons stored locally?

Lessons are saved as Markdown files in the ~/.claude/lessons/ directory, with a summary table automatically maintained in your ~/.claude/CLAUDE.md file.

Reflect: Continuous Learning Capture

Name: Reflect: Continuous Learning Capture
Author: Porkbutts

byPorkbutts

•

Productivity & Workflow

Captures and persists actionable lessons from conversation mistakes to improve Claude's performance in future sessions.

About

The Reflect skill introduces a structured feedback loop to your AI development workflow by enabling Claude to analyze its own mistakes and user corrections. When triggered, it scans the conversation to identify root causes, drafts a 'lesson learned' summary, and validates the insight with the user before persisting it to a local knowledge base. By maintaining a centralized index of these lessons in CLAUDE.md, the skill ensures that hard-won technical insights become permanent project context, preventing the recurrence of errors and accelerating future development tasks.

Key Features

Standardized persistence to local Markdown files with dated naming
Automated conversation analysis to identify mistakes and corrections
Mandatory user validation to ensure accurate root cause identification
Structured 'If we were starting over' guidance generation
1 GitHub stars
Centralized activity-based triggers in CLAUDE.md for easy reference

Use Cases

Capturing specific configuration steps after a difficult environment setup
Persisting architectural decisions and preferred coding patterns for a project
Documenting library-specific anti-patterns discovered during debugging

About

Key Features

Standardized persistence to local Markdown files with dated naming
Automated conversation analysis to identify mistakes and corrections
Mandatory user validation to ensure accurate root cause identification
Structured 'If we were starting over' guidance generation
1 GitHub stars
Centralized activity-based triggers in CLAUDE.md for easy reference

Use Cases

Capturing specific configuration steps after a difficult environment setup
Persisting architectural decisions and preferred coding patterns for a project
Documenting library-specific anti-patterns discovered during debugging