Sessions
The Sessions page is your review list. Every time a coding sub-session closes — a single goal you were pursuing in a stretch of a Claude Code session — it lands here with a grade and a status. Come here to see what you actually finished, understand why some work counted toward your score and some didn’t, and open any sub-session to learn how to prompt it better next time. It shows your own work only.
Filters
Section titled “Filters”The filter bar narrows a long list down to what you’re after:
- Date range — the last 28 days by default.
- Search — matches on intent, model, language, or close reason, so you can jump to “the Python refactor from Tuesday.”
- Grade — A, B, C, D, and F chips, to see only your best or worst.
- Status — filter to one outcome, for example only the ones that were redone.
- Intent — the kind of work (a bug fix, a plan, a question).
- Excluded only — flip this on to see just the sub-sessions that didn’t count toward your score. It’s the fastest way to understand your Completion rate.
A count line shows how many are displayed. To keep things fast, only the newest 500 in a range load at once; if you hit that limit, narrow the range to reach the older ones.
Reading the table
Section titled “Reading the table”Sub-sessions are grouped by day — or by a merge group, when several were really one goal. Each row is one sub-session:
| Column | What it shows |
|---|---|
| Title | The goal of the sub-session, with a status glyph and the close reason beside it |
| Turns | How many back-and-forths it took |
| Grade | Its Prompt Grade, A–F — how well you prompted it |
| TTC | Time to complete — from the first prompt to the finished goal |
| LOC | Lines of code it changed |
| In score | Counted or Excluded — whether it counts toward your Completion rate |
| Closed | When it closed |
The status glyph (Crushed, Rework, Incomplete, Trivial, or Pending) and the close
reason (topic shift, /clear, commit landed, or still in progress) sit inside the
Title cell, not in their own columns.
The legend above the table explains the status a row can carry, and each one tells you something different about the work:
- Crushed — it reached its goal and wasn’t redone. This is the one that counts.
- Rework — real work, but it redid something from an earlier attempt — a sign a fix didn’t hold the first time.
- Incomplete — real work, but the goal wasn’t reached (or wasn’t stated clearly enough to judge).
- Trivial / Pending — too small to score (a quick question, a one-line change), or not judged yet — one shared status covers both.
- Excluded from score — set aside for another reason, such as an outcome that couldn’t be judged or a merge into another sub-session.
Opening a sub-session to improve it
Section titled “Opening a sub-session to improve it”Click any row to open the Sub-session details panel. Badges at the top show its status, whether it was Crushed, its grade, and its language. Below them:
- Why this status — the checks the sub-session passed and failed, so you can see exactly why it counted or didn’t.
- How to improve — concrete suggestions for the prompts that fell short.
- Prompts in this sub-session — the prompts that made it up, so you can revisit what you actually wrote.
- Substance — whether it cleared the floor, with its turns, net lines of code, and mutating tool calls.
- Facts — the findings behind the verdict: crushed, goal completed, rework, model, rubric judgement, outcome check, and merge group.
- Identifiers — the sub-session ID, with a copy button.
If a grade looks wrong, open This grade looks wrong — pick a reason (the goal was misjudged, the task difficulty wasn’t accounted for, context was missing, or something else) and submit it for another look.