See this week's training recap

Sunday evening. The week is closing. You want the one-card recap — sessions hit, volume across the week, PRs landed, days missed. Whatever you're going to think about Monday morning starts from this read.

What the agent needs to hear

Behind your sentence, the agent works backwards to two pieces: which week and what view. The first defaults to this week, Monday through today — the cycle in progress. Say « last week » and it shifts to the previous completed Monday–Sunday instead.

The second defaults to the full recap card: sessions completed, total volume across the week, PRs landed, days missed, routine adherence. « How did training go this week » suffices. If you want a specific window — « the deload week », « two weeks ago » — name it.

What the weekly card covers

The card packs four reads into one surface. Sessions completed is the raw count alongside planned-vs-actual — four out of five hit, one missed Wednesday. Total volume sums the working sets across every session in the window, so you see the week as one number, not as a stack of session pages.

PRs landed lists what moved this week, or says « no PRs » when nothing did — silence is information too. Routine adherence is the texture: which planned days you hit, which you missed, which you substituted (« Friday’s leg day moved to Saturday »). The agent reads from the routine plan, not from session count alone, so a session under the wrong routine doesn’t count as adherence.

How this week reads in context

A weekly card in isolation is just a number. Its job is to read against the recent past — that’s where the signal lives. By default the agent surfaces a delta versus the prior completed week (« +2 sessions vs last week », « -8% volume »). The delta is what tells you whether the week was up, down, or flat.

For a longer view, ask for the week « vs last month’s average » or « vs the average of the last four weeks ». For a custom span, name two windows and the agent compares them directly. The weekly card is the entry point; the comparison is what turns it into a decision.

When the agent gets it wrong

Wrong window is the most common miss. On a Sunday « this week » reads as the cycle that’s closing; mid-week the phrase is ambiguous and the agent may pull a partial week as if it were complete. Pin it: « Monday May 5 through today », or « last completed week ».

Missing sessions are the silent miss. If a session was logged under a paused or archived routine, it counts toward volume but reads as off-plan in adherence. Ask « show me what’s tagged off-plan » and the agent surfaces the orphans — usually one or two sessions that need re-tagging.

The third miss is the delta. If last week was a deload, comparing this week against it reads as a big jump that isn’t real. Tell the agent « exclude the deload week » or « compare to the last non-deload week » and the baseline shifts to something honest.

What makes the recap worth reading

Three things decide whether this weekly read tells you something useful: the window matches your training cycle (the week you actually trained, not an arbitrary Mon–Sun if you’re on a 5-day rotation), the comparison is to a fair baseline (don’t read a normal week against a deload as « regression »), and missing sessions are surfaced as missing, not zeroed (a skipped Friday should be visible, not silently averaged out).

Read the recap, name the comparison, and next week’s plan adjustment writes itself.