Accountability ledger

The tab reads the team event store for a 30-day window and opens with two scalar cards:

• Money-saving outcomes: the count of measured outcome events whose recorded benefit was greater than zero.
• Dollars saved: the sum of the benefit recorded on those outcome events (action.outcome), formatted as USD.

Both are read from what your team's desktops already stream, scoped to the repos you can see; personal repos never roll up. The tab captures nothing new. When a window holds more events than one read returns, the totals count the most recent 5,000 and the page says they are a floor.

Below the cards, a table lists one row per day that carried a measured saving, newest first: the date, what was saved, what was spent that day, and the running total. A day with no measured saving is not a row, so an idle month does not read as a busy one. The table appears only once at least one day qualifies.

The last block on the page comes from somewhere else: what your team sent with repoops bench --contribute. That command measures one repo on one developer's machine (git blame over recorded defects, the local brain's lessons, tracked session costs) and sends thirteen aggregate fields: the month plus twelve headline numbers, no file paths, no session ids, no lesson text. The block shows attribution accuracy, the recurrence-prevention rate and the dollars the run measured, plus a row per month once more than one run has been contributed; a second run in the same month replaces that month's row. It stays apart from the totals above on purpose: folding a local measurement into a hosted rollup would give you one number nobody could trace to a source. When nobody has run the command, the block says so and prints the command rather than a zero.

The stored team events carry only per-day totals, so the hosted side has the counts and dollars but not the individual loops behind them. The full per-story chain (a session shipped a PR, it introduced a defect, and a lesson now guards it) stays in the desktop app over the local accountability graph. The hosted ledger points at that chain rather than reproducing it.

Under the totals, the tab links to the two brain-store surfaces that hold the knowledge doing the guarding: Lessons (the lessons that keep a defect from recurring) and Decisions (the decisions behind them). The totals tell you the measured effect; those surfaces tell you what stands between your team and the next incident.