workweavelisted

# yc:workweave Turn data agents-cli already keeps locally into a decision-useful engineering-intelligence report. This is a composed view over existing engines, not a new analytics pipeline. ## 1. Resolve scope Require a Git repository. Resolve the canonical project key from the origin repository name so linked-worktree directory names do not split one project into several dashboards. ```bash ROOT=$(git rev-parse --show-toplevel) ORIGIN=$(git remote get-url origin) PROJECT=$(basename "${ORIGIN%.git}") DATE=$(date +%F) OUT="$ROOT/.agents/artifacts/$DATE" WINDOW=30d mkdir -p "$OUT/workweave-data" ``` Parse `$ARGUMENTS` exactly: - `--since <window>` controls time-windowed engines; default `30d`. - `--since all` uses all history where supported. For an engine that does not accept `all`, derive the earliest timestamp from the session census and pass that ISO date. - `--no-open` renders and inspects without opening the user's browser. - Reject unknown arguments with the accepted forms instead of silently ignoring them. Set `WINDOW` to the parsed `--since` value when supplied. The session census is always all-time for `PROJECT`, independent of the chart window. ## 2. Refresh the local index and collect once Do not read transcript files directly. The bundled collector scans registered harness roots, updates changed sessions in the local index, paginates through every row, strips identity, paths, session IDs, topics, and recovery metadata in memory, then runs the independen