The scoring model

Prism’s headline metric is the Prompt Score — how well you direct the AI, expressed as a single 0–100 number with a letter grade:

Prompt Score = Average Prompt Grade (APG), 0–100 + letter grade

Every prompt is graded by seven rubric checks applied across eight intent classes and rolled up to a letter grade. Averaging those per-prompt grades gives the Prompt Score you see at the top of the dashboard.

Supporting mechanics

The Prompt Score sits on top of a sub-session model and an outcome check that together supply the supporting metrics:

• Sub-sessions. A Claude Code session is split into goal-arc sub-sessions — the unit Prism scores. Each also gets an LLM-generated title and 1–2 sentence summary that appear in the dashboard.

• Completion rate. The share of scored sub-sessions that crushed their goal:

  Completion rate = crushed sub-sessions ÷ scored sub-sessions

  A sub-session crushes when it clears the substance floor, an LLM outcome judge confirms it reached its goal under the criteria for that kind of task, and no later work undoes it. Completion rate is a supporting outcome metric, not the headline.

• Three supporting metrics. Speed (CSPW + TTC), Skill (Skill Index 0–100, tier Novice → Elite), and Token Usage (Crush Weight) break the picture down further.

For the full mechanics, see the dedicated section: Prompt Score.

• Algorithm overview