Ask if you're recovered enough to train hard

Wednesday morning. You did legs Monday and your hamstrings still hum a little. The plan says heavy squats today. Before you commit to the warm-up, you ask the agent for the readiness read — workload, streak, sleep if logged, the composite verdict.

What the agent needs to hear

Behind your sentence, the agent works backwards to two pieces: the session you’re contemplating and how detailed an answer you want. The session type sets the threshold — « recovered enough » for an easy zone-2 hour is a different bar than « recovered enough » for a heavy squat day. The detail level decides what comes back: composite score with the inputs that drove it, by default.

« Readiness? » with no context gives you a general snapshot — the agent picks a moderate-session threshold. « For a heavy squat day » narrows the bar. Naming the session is the cleanest way to get an answer that maps onto the decision you’re about to make.

What goes into the score

Four streams feed the read. Recent workout load weighs your volume and intensity over the last 5–7 days, with recency tilting the average — Monday’s heavy session counts more than the same session ten days back. Active streak shifts the threshold: longer streaks of consecutive hard days raise the bar, because diminishing returns are real. Session frequency reads the pattern, not just the total — back-to-back hard days look different than hard/easy/hard.

Logged wellness rounds it out when available: sleep entries, RPE on recent sessions, post-workout assessments. If a stream is silent — no sleep logged this week, no RPE on Monday — the agent surfaces which inputs were missing, so you know what’s driving the read and what’s being inferred.

What « readiness » actually answers

Readiness is a prior, not a license. A green read doesn’t promise the session will go well; a yellow read doesn’t mean you have to back off. The score is the Bayesian starting point: given your recent loads and signals, the probability your hard day lands cleanly.

The decision is still yours during the warm-up. If your first work-up sets feel wrong, the green read doesn’t override that. If the warm-up moves like silk, the yellow read doesn’t force a deload. Use the readiness check to inform the start of the session, not to replace the felt sense you’ll have once you’re under the bar.

When the agent gets it wrong

Three patterns recur. High score, you feel awful — usually a missing input. Sleep wasn’t logged this week, or Monday’s RPE never made it in, and the agent’s working blind on the wellness stream. Log what’s missing and re-ask.

Low score, but the last 5 days were easy — the agent may be over-weighting an older hard session. Narrow the window: « readiness based on the last 3 days only ». Wrong session-type threshold — you asked « readiness » in the abstract, the agent assumed heavy when you meant zone-2 cardio, and the read came back conservative. Name the session and the bar resets.

What makes the read worth trusting

Three things decide whether this readiness read informs your day usefully: the inputs cover the relevant window (5–7 days for most cases; shorter when you’ve recently changed your training pattern), the threshold matches the session you’re actually planning (heavy squats and easy cardio are different recovery asks), and the signal is treated as a prior, not a verdict (the score informs the start; the warm-up tells you the truth). Read the score, set today’s opening intensity from it, and let the bar tell you the rest.