Session Reflection Guide

Learn from your AI coding sessions by capturing what worked, what didn't, and patterns to reuse.

Purpose

Session reflection is meta-analysis - focusing not on what was built, but on HOW the work was done:

• Capture effective patterns - What approaches worked well?
• Document lessons learned - What would you do differently?
• Build personal knowledge - Accumulate tips and tricks over time

When to Reflect

• After complex multi-step tasks
• When discovering new effective patterns
• After significant debugging sessions
• When a session went particularly well (or badly)
• Before ending a long work session

Using the Reflect Prompt

This repo includes a /reflect prompt for generating structured reflections:

/reflect api-migration

Or let it auto-derive a name:

/reflect

The prompt creates a markdown file at .copilot/reflections/<date>-<name>.md.

Reflection Structure

What Went Well

Identify effective techniques:

• Problem-solving approaches that worked
• Tool usage that was particularly helpful
• Communication patterns that produced good results
• Planning strategies that paid off

What Went Wrong

Document issues for future avoidance:

• Approaches that didn't work or caused delays
• Tool misuse or inefficiency
• Unnecessary iteration or backtracking
• Context management problems

Lessons Learned

Extract actionable insights:

• Each lesson should inform future behavior
• Be specific about what to do differently
• Include the "why" behind the lesson

Tips & Tricks

Capture useful patterns discovered:

• Specific prompting techniques
• Workflow shortcuts
• Configuration discoveries
• Tool combinations that work well

Example Reflection

# Session Reflection: API Migration

**Date**: 2026-02-05
**Session Goal**: Migrate REST API endpoints to new versioned structure

---

## What Went Well

- Breaking migration into phases (routes → handlers → tests) kept scope manageable
- Using Agent mode for bulk file moves with pattern matching
- Keeping a checklist in _plans/ prevented missing endpoints

## What Went Wrong

- Started without checking existing test coverage - had to backtrack
- Tried to migrate everything in one conversation - context got confused
- Forgot to update API documentation until the end

## Lessons Learned

1. **Check test coverage first**: Before any refactor, know what's tested. Saves debugging time.
2. **One conversation per phase**: Don't mix migration phases in same context.

## Action Items

- [ ] Add "check test coverage" to personal migration checklist
- [ ] Create prompt template for API migrations

## Tips & Tricks

- **Tip**: Use `@workspace find all routes` to get comprehensive endpoint list
- **Tip**: Agent mode + explicit file patterns = reliable bulk operations

---

*Generated by `/reflect` prompt*

Building a Knowledge Base

Over time, reflections accumulate into a personal knowledge base:

.copilot/reflections/
  2026-02-01-auth-debugging.md
  2026-02-03-test-refactoring.md
  2026-02-05-api-migration.md

Periodically review reflections to:

• Identify recurring patterns (good and bad)
• Extract tips into instruction files
• Create prompts for common workflows

Generalization Opportunities

If a reflection reveals a repeatable pattern, consider creating:

Pattern Type	Create
Repeated workflow	Custom prompt
Specialized task	Agent configuration
Multi-step process	Skill with reference files
Common instructions	Add to instruction files

Only generalize patterns that genuinely recur across multiple contexts. Premature abstraction adds complexity without benefit.

Summary

Step	Action
Trigger	End of significant session
Generate	Use /reflect prompt
Review	Read and refine the output
Store	Keep in .copilot/reflections/
Apply	Use insights in future sessions
Generalize	Create prompts/instructions from patterns