Two quick checks that catch a misunderstood assignment before it wastes days

Two understanding checks that run before a person starts a new task. The first scores the notes they took while reading the brief against the actual key points of that brief. The second asks them, right before they start, to explain the goal of the task back in their own words. Both results feed into the system's running read of how clearly this person understood the assignment going in.

Misreading a brief is expensive: it can mean days spent building the wrong thing before anyone notices. Catching that misunderstanding before work starts is far cheaper than catching it after. The first check runs while someone is still reading the brief, scoring how well their notes match its key points. The second runs right before they dive in, asking them to explain the goal in their own words, so the system can tell whether they actually understood it. If that explanation is vague, the system knows to intervene instead of letting them start already off track.

Both checks run as real-time endpoints backed by a lightweight AI agent. The note check scores each note the person wrote against the canonical key points from the task brief and returns what matched, what didn't, and an overall understanding percentage. The explain-it-back check is intentionally kept thin, with no extra coaching wrapped around the model, so it can't soften its own verdict when someone gives a vague answer. Both checks write their result into the person's running skill record, so later, more thorough reviews know how clearly this person started the task and can weigh their independence and quality against that starting point.

Sole author of both checks.