8. The Done panel

Done is what shipped, grouped by the iteration it shipped in. The bands are facts: once a story was accepted in week N, that's where it lives forever. agilemarkdown doesn't re-bin Done by current velocity or any other moving target.

8.1 What shows up

The Done panel filters every story where:

1. status is accepted, and
2. the accepted timestamp falls in a prior iteration (not the current one).

Stories accepted in the current iteration stay in the Backlog band above the in-flight line. The moment Monday's tick advances the current iteration number, those accepted stories silently roll into Done without any batch job or archive step; the row and the file are the same, but the filter result is different because now advanced.

VS Code Shell MCP

$ am show done
Done (stories)   18 stories

── Iteration 1376  Mon May 11 ──
   1. ★  Wave once at the moon         accepted    1p
   2. ★  Address the very first letter accepted    3p
   3. ★  Hum a small tune over the seed accepted   2p

── Iteration 1375  Mon May 04 ──
   4. ★  Water the seed at exactly dawn accepted    3p
   5. ★  Toss the last crumb            accepted    1p
   6. ★  Speak to the seed softly       accepted    1p

   ... and so on for prior iterations

done_list returns the same rows, newest first, with the accepted ISO timestamp on each. The client groups them by iteration in the UI; the API just hands you the list.

Full envelope

{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "done_list",
    "arguments": { "backlog": "stories" }
  }
}

Response: result.structuredContent.items is an array sorted descending by accepted timestamp. Each row has the full OrderRow shape, same one as priority_list, so a client can render a Done card identically to a priority card.

8.2 The Friday retro

am retro reads the same accepted-story timeline shown above and folds in this iteration's headline numbers: velocity, volatility, cycle time, rejection rate. Print it before the team sits down on Friday and there is a starting agenda. Anything the team wants to remember coming out of the conversation goes into .am/learnings.md via record-learning. The retro view prints those back under "Recent learnings", so the next Friday's retro starts by reading them.

$ am retro
Retro summary

  velocity:        7 pts
  volatility:      22%
  cycle time:      21.2d (median)
  rejection rate:  0% (latest, target band 5-15%)
  accepted total:  21
  rejected total:  1

Three questions:
  1. What worked?
  2. What did not work?
  3. What changes for next iteration?

Recent learnings:
  - On the morning of 2026-05-07: yelling at the seedling
    did not help. We have agreed to try kindness instead.
  - On the afternoon of 2026-05-14: the cookies were better
    when shared. We will remember this.

$ am record-learning "Pointing the gnarly one first kept the rest honest."
- 2026-05-28: Pointing the gnarly one first kept the rest honest.

{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "record_learning",
    "arguments": {
      "note": "Pointing the gnarly one first kept the rest honest."
    }
  }
}

What just happened

• priority_list and done_list are mutually exclusive filters over the same underlying data. A story is in exactly one of them at any moment, decided by the timestamp.
• The iteration bands in the Done panel are derived from each row's accepted date, not from "current iteration minus N". A story shipped in May 4's week stays under "May 4" forever.
• Release markers shipped in the past appear here too, anchored to their accepted (ship) date.
• am retro reads the same accepted-story data and adds the iteration's numbers, so the Friday meeting has an agenda without a separate setup step.
• Learnings recorded with record-learning accumulate in .am/learnings.md and surface in the next retro's "Recent learnings" section.

Next: Find things three ways →