# End-of-Session Hygiene Checklist + Summary Prompt

**Source article:** [End-of-Session Hygiene: What to Capture Before Context Resets](https://tooearlytosay.com/research/methodology/end-of-session-hygiene/)
**Pattern:** Five-minute capture, four categories, one prompt.

Before closing a productive AI-assisted session, we spend five minutes capturing what would otherwise vanish overnight. Documentation is comprehensive and polished. Capture is quick, rough, and intended for future-us and future-AI. The distinction matters.

---

## The five-minute capture (four categories)

Append to a session log in `<your-project>/docs/session-log.md` or wherever our team keeps dated entries.

### 1. What changed

```
## Session: <YYYY-MM-DD>

- <one-line description of each meaningful change>
- <file paths if useful>
- <bug fixed, function refactored, edge case handled>
```

### 2. Why it changed

```
<plain-language rationale for non-obvious choices>
<the constraint or deadline that drove the trade-off>
<the discovery in production logs / data that motivated the fix>
```

### 3. What is next

```
Next session:
- <concrete first thing to pick up>
- <follow-up implied by today's work>
- <open question to investigate>
```

### 4. What we tried that did not work

```
Rejected approaches:
- <approach 1>: <reason it failed>
- <approach 2>: <reason it failed>
```

The fourth category is the one we are most tempted to skip and the one that saves the most future time. Without it we will re-explore dead ends we already ruled out.

---

## Where to put it (hybrid placement)

Different types of context belong in different places.

- **What the AI needs tomorrow** → update the "Current State" section of `CLAUDE.md` (the agent reads this).
- **What we might need in a month** → append to `docs/session-log.md` (human reference).
- **What future maintainers need** → inline comments at the decision site in the code.

---

## The session summary prompt

When we are ready to end the session, ask the agent to help capture:

```
Before we end, summarize what we accomplished, what decisions we made
and why, what we tried that did not work, and what the next steps are.
Format it for the CLAUDE.md "Current State" section.
```

The agent has perfect recall of the session and can extract the key points faster than we can. We review, edit if needed, and paste into the appropriate locations.

This takes two minutes and saves twenty minutes tomorrow.

---

## Signs we skipped capture

If we notice any of these at the start of a new session, the previous session's hygiene fell off:

- We start with "what were we working on yesterday?"
- We re-implement something we already built.
- We re-explore a dead end we already ruled out.
- We cannot remember why we made a particular decision.

Each of these costs more time than the five-minute capture would have taken.

---

## What not to capture

Capture is not about recording everything. It is about recording what is not obvious from the code itself.

**Skip:**
- Changes that are self-evident from the diff.
- Decisions already documented in code comments.
- Standard patterns that need no explanation.

**Capture:**
- The "why" behind non-obvious choices.
- Context that influenced decisions (deadlines, constraints, trade-offs).
- Plans that exist only in our heads.
- Approaches we rejected and the reasons.

The test: if we read just the code, would we understand this? If yes, skip. If no, capture.

---

*Template extracted from work published on [Too Early To Say](https://tooearlytosay.com).*
*Licensed MIT. Victoria Cholette, 2026.*