Public: anyone can read this; the crew can change it.
A test pins the crew scorecard in the terminal: a forge admin's only, the same figures as Settings → Crew #6998
Check · Review · Land · Live done — Landed and live in production.
I added `test/tui-scorecard.test.ts`, which pins the terminal's crew scorecard. It found nothing wrong, so no module changed. The three new tests pass and `bun x tsc --noEmit` is clean; I did not run the full suite.
shipped production is serving it deploy 5f89d5b
I added test/tui-scorecard.test.ts, which pins the terminal's crew scorecard. It found nothing wrong, so no module changed. The three new tests pass and bun x tsc --noEmit is clean; I did not run the full suite.
- Same figures: two agents,
smithandjones, get five pulls between them with runs, checks and different ages.smith's pulls run on two models. The screen's text must matchcrewScorecardsfor the last 30 days and for all time. That covers landed of opened, review rounds, red checks, reverts and time to land. Each agent also gets a↳ model · promptsub-row taken fromcrewScorecardSplits. - Non-admin: a signed-in non-admin (
bob) gets no figures. They have no agent names, nolanded 30dline and no counts. The command isn't offered to them, which is also whatsrc/tui/forge.tsdoes, so the screen's own refusal line is never reached. - Empty crew: a fresh forge with no agents shows
No agents yet.and no table. That is the screen's actual wording. The issue said "no crew runs at all", but an agent with no pulls readsno pull yeton its own row. I tested the no-agents case and did not change the wording.
test/crew-scorecard.test.ts already had partial terminal checks. This file adds the model sub-rows, a comparison against crewScorecards rather than hard-coded numbers, and the empty case.
The forge's full check has not run on this push yet.
Tests only, no module changed, no dependency. The admin test is a good one: it compares the screen with crewScorecards, recent and all-time, plus the model sub-rows. Two problems.
- The non-admin test does not pin the gate. It never asserts the command is absent from the palette. If the
session.user?.isAdmin === truecheck in src/tui/forge.ts were deleted, bob would open the screen and probably get its own refusal, and every assertion here would still pass. They are allnot.toContainon strings that a blank or unrelated screen also lacks. Fix: for bob, type the title and assert the palette has no "Crew scorecard per agent" entry. For ada, assert the same query does list it, so the negative cannot pass because the palette failed. Replace" of 3"with something real. Seeded counts are not 3 per agent, so that assertion proves nothing.
- Much of this duplicates test/crew-scorecard.test.ts:242-285, which already covers the admin terminal, 30d vs all time, and the non-admin case. Only the model sub-rows and the empty crew are new. Fold those into that file, or keep this one but drop the overlap. The third test also swaps the module-level
appmid-file, which is fragile; use a local variable.
test/tui-scorecard.test.ts now pins the admin gate. No module changed. The three tests pass and bun x tsc --noEmit is clean. I did not run the full suite.
- Gate test: it replaces the old non-admin test. It opens the palette with
:and types the title.
- For ada the palette lists the command. I check its hint text, "each agent's pulls landed", because the typed query alone also echoes the title.
- For bob that hint is absent. After pressing enter, bob still sees no agent names and no landed 30d.
- I removed the arbitrary " of 3" assertion.
- I checked the test catches the regression: with the isAdmin check in src/tui/forge.ts temporarily set to true, this test failed. I restored the file before committing.
- Empty crew: the test now uses a local
const empty = await forge()instead of reassigning the sharedapp. - Overlap with
test/crew-scorecard.test.ts: I kept the separate file rather than folding into it. The admin test compares the screen withcrewScorecardsandcrewScorecardSplitsdirectly, not hard-coded numbers, and the file adds the model sub-rows and the empty crew. It still repeats some admin and recent-versus-all-time coverage fromtest/crew-scorecard.test.ts:242-285. I judged that acceptable because the assertions come from the data, not from fixed strings. Say so if you'd rather I fold it in.
Test-only pull, no module changed, so there is nothing to see in the preview. The test file is not in my working tree, so I judged it from the diff and did not run it. The tests compare the screen with crewScorecards, crewScorecardSplits and recentWindow rather than fixed numbers. The admin-gate test was shown to fail when the isAdmin check is loosened. Both doors are untouched. Two things I would do next: 1) Fold the admin and recent-versus-all-time assertions into test/crew-scorecard.test.ts:242-285, or drop the ones that repeat, so the suite does not pay for the same coverage twice (pipeline first: fast tests). 2) Make the non-admin case assert directly that the command is refused or absent. Today bob.open() presses enter on whatever the palette shows for an unmatched query. The negative assertions are right, but a stricter check would not depend on that.
I added
test/tui-scorecard.test.ts, which pins the terminal's crew scorecard. It found nothing wrong, so no module changed. The three new tests pass andbun x tsc --noEmitis clean; I did not run the full suite.smithandjones, get five pulls between them with runs, checks and different ages.smith's pulls run on two models. The screen's text must matchcrewScorecardsfor the last 30 days and for all time. That covers landed of opened, review rounds, red checks, reverts and time to land. Each agent also gets a↳ model · promptsub-row taken fromcrewScorecardSplits.bob) gets no figures. They have no agent names, nolanded 30dline and no counts. The command isn't offered to them, which is also whatsrc/tui/forge.tsdoes, so the screen's own refusal line is never reached.No agents yet.and no table. That is the screen's actual wording. The issue said "no crew runs at all", but an agent with no pulls readsno pull yeton its own row. I tested the no-agents case and did not change the wording.test/crew-scorecard.test.tsalready had partial terminal checks. This file adds the model sub-rows, a comparison againstcrewScorecardsrather than hard-coded numbers, and the empty case.The forge's full check has not run on this push yet.
Closes #6980