Free AI Skill
After every session,
know what to keep.
Run a quick debrief after any AI work session. Assess what you built, what went sideways, and whether anything should become a reusable prompt, a standing rule, or an automation. Works in any AI tool.
Any AI tool
5 minutes
Reusable prompts
No prompting needed
What it does
🔍
Review
Reconstructs the session
Looks at what was built, corrected, and discovered — main tasks, course changes, friction points, and any repeatable workflows that emerged
❓
Check
Runs 4 questions
New reusable prompt? New rule? Worth automating? Worth saving for later? If yes, it drafts the thing immediately
⚡
Build
Drafts what's warranted
Writes out the reusable prompt, rule, or workflow description — ready to save wherever you keep your prompts
📋
Summary
Reports what was kept
Outputs a clean summary of what was captured and why — so you know your system got smarter
What gets captured
The debrief checks four things and drafts the output immediately if the answer is yes.
Reusable Prompt
Did we do something repeatable? Drafts a clean, copy-paste prompt you can save and reuse next time.
Standing Rule
Did a correction come up? Writes it as a permanent instruction so you don't re-explain it every session.
Automation Candidate
Did we do something repetitive? Describes what to automate and how, so you can decide if it's worth the setup.
Context to Save
Did we learn something worth carrying forward? Captures preferences, environment details, or context worth including next time.
What you need
- Any AI tool — ChatGPT, Claude, Gemini, Cursor, or any chat interface
- A summary of what you did — bullet points of the session work, what broke, and what you corrected. 5-10 bullets is enough.
How to do it
Step 1
Run the debrief
At the end of any meaningful AI work session, paste this prompt into your AI tool with a quick summary of what happened. You don't need to be precise — bullet points are fine.
Here's a summary of what I did in this AI session:
[PASTE YOUR BULLET POINTS — e.g.:
- Drafted a cover letter for a Director of Demand Gen role at Stripe
- Had to re-prompt 3 times because the tone was too formal
- Added a rule: "lead with results, not setup"
- Also searched job boards and found 4 leads to add to my tracker]
Now run a session debrief. For each of these four questions, answer yes or no and explain briefly:
1. REUSABLE PROMPT: Did we do something step-by-step that I'll want to repeat? If yes, draft the prompt I should save.
2. STANDING RULE: Did I have to correct something that should become a permanent instruction? If yes, write the rule I should save.
3. AUTOMATION: Did we do something repetitive enough to be worth automating? If yes, describe what to automate and roughly how.
4. CONTEXT TO SAVE: Did we learn something about my preferences or setup worth carrying into future sessions? If yes, write what should be saved.
For each yes, draft the output immediately. Don't ask me if I want it — just produce it.
Takes about 5 minutes. Save any outputs — reusable prompts, rules, or notes — wherever you keep your AI reference material.
Step 2
Save what the debrief produces
If the debrief surfaced a reusable prompt, a standing rule, or a workflow worth automating, save it somewhere you'll actually use it. A Notion doc, a text file, or a dedicated prompts folder all work.
Take the outputs from the debrief above and format them for saving:
For any reusable prompt: format as a titled, copy-paste-ready block with a one-line description at the top.
For any standing rule: format as a short instruction I can paste at the top of future sessions with this AI.
For any automation candidate: write a 3-5 step description of what it would do and what tool would handle it.
Label each section clearly. I want to be able to drop these into my prompt library or notes without editing.
Claude Code users: the /debrief command writes skill files, edits CLAUDE.md, and creates scheduled tasks automatically — no manual saving.
⚡ Claude Code Shortcut
Power User
Using Claude Code?
Skip all of the above. I built this as a native Claude Code skill — one command reviews the session, runs the four-question check, and automatically writes skill files, edits CLAUDE.md, or creates scheduled tasks. No copy-paste, no manual saving.
In Claude Code, /debrief actually writes the files. In other AI tools, the prompts above output the text — you save it yourself. Both get you the same thinking; Claude Code just automates the filing.
Built by
◈
Clay Parker
Sr. Demand Gen & Lifecycle, Zapier · GTM & Revenue Marketing · AI-native builder
Related skills