Public: anyone can read this; the crew can change it.
/review: one quiet column where each item says what, why it needs you, the crew's recommendation and the evidence #4337
/review now lists the asks as one read-only column, newest first. I haven't loaded the page in a browser or timed it with 50 seeded asks.
stopped parked by the crew deploy e420fe2
2 earlier comments
/review now lists the asks as one read-only column, newest first. I haven't loaded the page in a browser or timed it with 50 seeded asks. It did return 200 under test with three asks, and bun x tsc --noEmit, bun scripts/design-lint.ts and test/design-lint.test.ts, test/design-audit.test.ts and test/review-page.test.ts all pass. I did not run the full bun test.
Signed-out readers are redirected to sign-in. Each ask is a card showing:
- What: the title (a link to the pull or issue), the repository, the noun "pull request" or "issue", and the age.
- Why: the
whysentence. - Recommendation: the crew's recommendation and its reason, as a badge plus reason. It reads the crew's
Recommendation:line from the thread, so an issue gets one too. With none, the card says "The crew has no recommendation." - Evidence: diff stat, check result, preview link, and before/after links.
It reuses decisionsFor from src/web/studio/asks.ts, so the rows are exactly what /asks shows. Server-Timing names session, asks and render; asks covers reading the asks, their evidence and the diff stats.
I added no class, colour, style= or off-scale size. /review is in docs/design/audit.md and the count there is now 62. test/review-page.test.ts seeds a handed-over pull, a 30-hour pull and a parked issue; it finds exactly those three rows, each with its why and recommendation, and leaves out a young crew pull.
Things to know when reviewing:
- Order: "newest decision first" sorts by when the pull or issue was last updated. The asks carry no decision timestamp, so this is my reading of the phrase.
- Cap: the column shows at most 50 asks.
- Diff stats:
gitruns once per pull, in parallel. That is where the 200 ms budget is most at risk. A pull whose diff can't be read shows no stat. - Before/after: these are links to the look pictures, not inline images, because there is no named image class. "Before" is the production look and "after" is the preview's. They appear only when the repository is an app with looks taken.
- Left out:
/asksis unchanged, and acting from the page is left for the next issue.
The first push got a 503 from the forge; a retry went through.
/review is now listed in app.looks in forge.json, which fixes the failing test "app.looks is the router's top-level pages minus the not-a-page list". That test requires every new top-level route to be either declared in app.looks or listed as not-a-page, and I had added the route without declaring it.
I ran test/forge-looks.test.ts and all 4 tests pass, including the one that was failing. I did not re-run the whole suite. In the check log, shards 5, 6 and 8 exited 143 without naming a test; I believe the fail-fast skip after the red shard caused that, but I haven't confirmed it. The forge's check on this push will show whether they pass.
I pushed the change to the pull.
Pull description
/review is a new page that lists the crew's asks as one read-only column, newest first. It is signed-in only; anyone else is sent to sign-in.
- Contents: each ask is a card with the title (linked), repository, noun and age, the
whysentence, the crew's recommendation with its reason, and the evidence. The evidence is the diff stat, check result, preview link and before/after look links. - Source: it reads the same asks as
/asksthroughdecisionsFor, which is unchanged. - Timing: Server-Timing names
session,asksandrender. I did not time it with 50 seeded asks, and the diff stat runsgitonce per pull, in parallel, so that is where the 200 ms budget is most at risk. - Design: it adds no class, colour or
style=, anddocs/design/audit.mdlists the page and counts 62 top-level pages. - Looks:
/reviewis inapp.looksinforge.json. - Tests:
test/review-page.test.tsseeds a handed-over pull, a stale pull and a parked issue. It finds exactly those three rows, each with its why and recommendation, and leaves out a young crew pull. - Choices:
- "Newest first" sorts by the pull or issue's last update, because the asks carry no decision timestamp.
- The column shows at most 50 asks.
- Before/after are links, not inline images, because there is no named image class.
- Left out: acting from the page (the next issue) and folding
/asksinto/review(the owner's call).
🤖 Generated with Claude Code
Parked by the crew: the check failed ("reserved names > every top-level page of the forge is a reserved name") and the decider saw no point in another try
/reviewnow lists the asks as one read-only column, newest first. I haven't loaded the page in a browser or timed it with 50 seeded asks. It did return 200 under test with three asks, andbun x tsc --noEmit,bun scripts/design-lint.tsandtest/design-lint.test.ts,test/design-audit.test.tsandtest/review-page.test.tsall pass. I did not run the fullbun test.Signed-out readers are redirected to sign-in. Each ask is a
cardshowing:whysentence.Recommendation:line from the thread, so an issue gets one too. With none, the card says "The crew has no recommendation."It reuses
decisionsForfromsrc/web/studio/asks.ts, so the rows are exactly what/asksshows. Server-Timing namessession,asksandrender;askscovers reading the asks, their evidence and the diff stats.I added no class, colour,
style=or off-scale size./reviewis indocs/design/audit.mdand the count there is now 62.test/review-page.test.tsseeds a handed-over pull, a 30-hour pull and a parked issue; it finds exactly those three rows, each with its why and recommendation, and leaves out a young crew pull.Things to know when reviewing:
gitruns once per pull, in parallel. That is where the 200 ms budget is most at risk. A pull whose diff can't be read shows no stat./asksis unchanged, and acting from the page is left for the next issue.The first push got a 503 from the forge; a retry went through.
Closes #3257