A check's run page shows its steps and live progress, and the runs list shows checks by default #1124

Pillar: P0 Trust (umbrella #586). Second half of #1102. It waits on #1123.

Intent. Looking deeper at a check works like looking at an Actions run: steps, progress and failures first, and the full tail below.

Done looks like.

• [ ] A check's run page shows its steps as rows (install, typecheck, tests), each with its state and duration.
• [ ] The tests step has a progress bar (files started out of the gate's count, passes and failures so far, the file running now), read from the same source as #1123.
• [ ] Failures pin to the top as they land, each with its first lines. The full log tail stays below.
• [ ] /:owner/:name/runs lists checks and deploys with the agent runs by default, with the three as filters. A check's row names the pull, the head, the state and how long it took.
• [ ] Tests for the step rows, the pinned failure and the default list.

What this leaves out of #1102, on purpose. Feed entries for every check start and finish. The owner asked to cut noise (2026-10-12), so a check reaches the feed only when it fails a pull he is waiting on, and that is a filter question for #1112.

Files. the run page under src/web/pages/, src/checks/summary.ts, the runs list, a test.

Risks. Layout shift as rows appear. Render all three step rows from the first paint.

Suggested: the same builder as #1123.

Parked by the crew: the check failed ("origins on the web > connecting with a token stores it and never echoes it") and the decider saw no point in another try (pull #2367 closed). Planner: split this smaller, or rewrite what done looks like; a person can also say what to try.

Parked by the crew: the check failed ("the code browser on the web > (unnamed)") and the decider saw no point in another try (pull #2373 closed).

Parked twice since a person last spoke here, so it is the planner's now. Every reason:

1. the check failed ("origins on the web > connecting with a token stores it and never echoes it") and the decider saw no point in another try (pull #2367 closed)
2. the check failed ("the code browser on the web > (unnamed)") and the decider saw no point in another try (pull #2373 closed)

The planner splits it, replaces it with a fresh issue, or defers it; a person can also say what to try. Nothing starts here until then, and the board lists it.

Split into #2381 (the runs list's default and filters), #2382 (step rows from the first paint) and #2383 (the progress bar and pinned failures). Both tries changed the list, the run page and the summary at once, and went red in test/web-origins.test.ts and test/web-code.test.ts. Each piece now touches one page and names those two suites to run.