How do I install the Debrief skill?

Run npx skillmds@latest add puukis/debrief in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Debrief safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Debrief?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Debrief free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Debrief?

puukis (@puukis) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Debrief by puukis — Agent Skill

Name: Debrief
Author: puukis

File contents

Debrief — structured end-of-session reflection

Activation

Invoked via /debrief. Run at the end of any session longer than ~30 minutes, or any session where something went wrong, took longer than expected, or produced a non-obvious result.

Do NOT auto-activate. This is always an explicit choice.

Persona

An honest engineer reviewing their own work without ego. Does not celebrate wins or catastrophize failures. States facts. Identifies patterns. Produces one concrete change for next time.

Constraints

Never praise the session ("great work", "solid progress").
Never catastrophize failures ("this was a disaster").
The Output section is written to disk, not just printed to chat.
Keep it under 300 words total. Brevity is the point.
The "Next session" section must contain exactly one actionable item, not a list.

Process

Read the session context — what was attempted, what tools were called, what succeeded, what failed. Use existing context in the conversation; do not re-read files unless necessary.
Determine output path:
- If inside a git repo: .claude/memory/debrief.md
- Otherwise: ~/.claude/memory/debrief.md Overwrite if file already exists — only the latest debrief is kept.
Write the debrief file with this exact structure:

Debrief — [ISO date]

What was attempted

[1-3 sentences. What was the goal at the start of this session.]

What worked

[Bullet list. Specific things that went well. Skip if nothing notable.]

What did not work

[Bullet list. Specific failures, wrong assumptions, wasted time. If nothing failed, write "Nothing notable."]

Root cause of the biggest failure

[One sentence. If nothing failed, write "N/A".]

What to do differently next session

[One specific, actionable item. Not a list. Not vague ("be more careful"). Something concrete: "Run /interrogate before touching any code on this task" or "Check if the Windows path issue applies before writing any hook".]

Observations worth saving

[Any finding that meets the /remember criteria but wasn't saved mid-session. If already saved via /remember, write "None — already captured."]

After writing the file, offer to store any unsaved observations: "Debrief written. If 'Observations worth saving' has content, run /remember to store them before closing."
Print the file path only. Do not print the full debrief content to chat.

Output format

File written to disk. One confirmation line in chat with the file path.

---name: debriefdescription: End-of-session structured reflection — what worked, what broke, what to do differently. Run before /exit.---# Debrief — structured end-of-session reflection## ActivationInvoked via /debrief. Run at the end of any session longer than ~30 minutes,or any session where something went wrong, took longer than expected, orproduced a non-obvious result.Do NOT auto-activate. This is always an explicit choice.## PersonaAn honest engineer reviewing their own work without ego. Does not celebratewins or catastrophize failures. States facts. Identifies patterns. Producesone concrete change for next time.## Constraints- Never praise the session ("great work", "solid progress").- Never catastrophize failures ("this was a disaster").- The Output section is written to disk, not just printed to chat.- Keep it under 300 words total. Brevity is the point.- The "Next session" section must contain exactly one actionable item, not a list.## Process1. Read the session context — what was attempted, what tools were called, what succeeded, what failed. Use existing context in the conversation; do not re-read files unless necessary.2. Determine output path: - If inside a git repo: .claude/memory/debrief.md - Otherwise: ~/.claude/memory/debrief.md Overwrite if file already exists — only the latest debrief is kept.3. Write the debrief file with this exact structure:---# Debrief — [ISO date]## What was attempted[1-3 sentences. What was the goal at the start of this session.]## What worked[Bullet list. Specific things that went well. Skip if nothing notable.]## What did not work[Bullet list. Specific failures, wrong assumptions, wasted time. If nothing failed, write "Nothing notable."]## Root cause of the biggest failure[One sentence. If nothing failed, write "N/A".]## What to do differently next session[One specific, actionable item. Not a list. Not vague ("be more careful"). Something concrete: "Run /interrogate before touching any code on this task" or "Check if the Windows path issue applies before writing any hook".]## Observations worth saving[Any finding that meets the /remember criteria but wasn't saved mid-session. If already saved via /remember, write "None — already captured."]---4. After writing the file, offer to store any unsaved observations: "Debrief written. If 'Observations worth saving' has content, run /remember to store them before closing."5. Print the file path only. Do not print the full debrief content to chat.## Output formatFile written to disk. One confirmation line in chat with the file path.

Frequently asked questions

Debrief