Agent skill

20-05-debrief

Write a full-fidelity post-mortem after completing development work. Use after any merge, feature completion, or significant work session.

Stars 163
Forks 31

Install this agent skill to your Project

npx add-skill https://github.com/majiayu000/claude-skill-registry/tree/main/skills/other/other/20-05-debrief

SKILL.md

20.05 Post-Mortem

Write a post-mortem after every significant piece of development work. This is NOT just for incidents — it's a full-fidelity account of what happened during the work so future agents and humans understand the real story.

Post-mortems MUST be committed to main. They are project history, not branch artifacts. If you're on a feature branch, switch to main (or finish your merge first via 20.04) before creating the post-mortem.

When to Write

  • After every merge (triggered by 20.04)
  • After completing a feature or significant refactor
  • After any work session where decisions were made, problems were hit, or lessons were learned

There is no threshold. Don't skip because "nothing went wrong" or "it was straightforward." Straightforward work still has decisions, context, and lessons worth capturing.

Create the Post-Mortem

  1. Make sure you're on main
  2. Create a tk ticket tagged post-mortem with the template below filled in
  3. Commit to main

Template

markdown
# Post-Mortem: <title>

**Date:** <date>
**Branch/Feature:** <branch or feature name>
**Duration:** <rough time spent>

## Goal

What were we trying to accomplish? One or two sentences.

## What Actually Happened

Factual replay of the work in sequence. Include:
- What was tried first
- What changed along the way
- Key timestamps or commits if relevant
- Links, quotes, error messages — concrete evidence

This section is calm and factual. No analysis yet.

## Decisions Made

For each significant choice:
- **Decision:** What we chose
- **Alternatives considered:** What else was on the table
- **Why:** The reasoning at the time
- **In hindsight:** Would we choose the same? Why or why not?

## Problems Hit

What went wrong or was harder than expected:
- What the problem was
- How it manifested (error messages, unexpected behavior, confusion)
- How it was resolved
- How long it took vs how long it should have taken

## What Worked

What went smoothly:
- Tools, patterns, or approaches that were effective
- Decisions that paid off
- Things to repeat next time

## What We'd Do Differently

Hindsight improvements — not blame, but honest reflection:
- Process changes
- Different technical approaches
- Better preparation or research
- Things that were over-engineered or under-engineered

## One Binding Improvement

Pick exactly ONE concrete, actionable improvement. Not five aspirational ideas — one thing that will actually get done.

- **What:** <specific change>
- **Who:** <person or agent responsible>
- **When:** <deadline or trigger>

Rules

  1. Commit to main. Post-mortems are project history. Never leave them on a feature branch or worktree.
  2. Write with full fidelity. Don't sanitize, don't summarize into nothing. The goal is that someone reading this in 3 months understands what really happened.
  3. Blameless. Describe what happened and why, not whose fault it was. People (and agents) make decisions under uncertainty — document the context, not the blame.
  4. One binding improvement. Many post-mortems produce a list of 10 action items that nobody does. Pick one. It must actually happen. If you're motivated to do more, great — but only one is required.
  5. Do it now. Don't defer "until later" — memory fades fast. The best post-mortem is written immediately after the work while context is fresh.
  6. Include the ugly parts. Dead ends, wrong approaches, wasted time, confusion — these are the most valuable parts for future readers. A post-mortem that only describes success is useless.

Expand your agent's capabilities with these related and highly-rated skills.

Didn't find tool you were looking for?

Be as detailed as possible for better results