score

name

score

description

Rate how well you prompted Claude in each session (optional, AI-graded). Runs the local prompt-health scorer over your ~/.claude transcripts and writes a 🟢/🟡/🔴 rating, a session type, and a one-line coaching note per session - shown in the dashboard. Needs your own Anthropic API key; runs entirely on your machine, and only the ratings (never your prompts) are ever stored or, if you host it, pushed. Run when the user types /time-tracker:score, or says "rate my prompts", "score my sessions", "how well did I prompt".

Score - grade your prompting (optional)

Adds a prompt-health rating to each session: a 🟢/🟡/🔴 grade, the session type (explore / build / debug / mixed), and a short coaching note. Ratings are calibrated to session type - an exploratory back-and-forth isn't penalised for meandering the way a build session would be. Results are cached locally and surfaced as badges in the dashboard.

Requires an Anthropic API key (the user's own). Scoring runs on this machine against local logs; only the derived rating/type/note are saved - never the prompts or code.

How to run

1. App dir: APP_DIR="${CCTI_APP_DIR:-$HOME/.claude-time-tracker}". If missing, run /time-tracker:setup first.
2. API key. Ensure ANTHROPIC_API_KEY is available - in the environment or in $APP_DIR/.env.local. If the user doesn't have one, point them to console.anthropic.com → API keys. Never ask them to paste it into the chat; have them put it in .env.local themselves.
3. Run the scorer:

   cd "$APP_DIR"
   npm run score-sessions -- --new     # only sessions not yet scored
   

   Use -- --all to score everything, -- --rescore to redo, -- --limit N to cap a run.
4. Report how many sessions were scored and remind the user the badges appear on next dashboard reload (/time-tracker:launch if it isn't running).