Runs and audit pages match the plain landing: no gradient wash, no blur

Claude Code·HighnessAtharva·yesterday·23hr 38min·10 Checkpoints·55 file changes·+1394/-1061

DOING A TEST FOR THIS

<pasted_content id="b4ef"> Goal: test the quickstart at https://help.altimate.ai/code/getting-started/quickstart/ the way a new developer would, fix any step that breaks, and prove the fix.

  1. Run docker info. If it fails, tell me what is missing and how to fix it. On Windows, use Docker inside WSL and set FIRSTRUN_DOCKER. Stop until it works.
  2. Run pip install git+https://github.com/HighnessAtharva/firstrun.
  3. Check that CLAUDE_CODE_OAUTH_TOKEN is set. If not, ask me to run claude setup-token and set it. Never print a secret.
  4. Run firstrun https://help.altimate.ai/code/getting-started/quickstart/. While it runs, tell me in plain words which repo it uses and which step it is on.
  5. Read runs/<id>/report.md. Tell me in 3 to 6 short sentences: which step broke, the error, the class Jev gave it, the fix, and whether a fresh container verified the fix.
  6. If the fix is verified and patches the docs, show me the diff and the PR body. Ask "Open this PR on <repo>?" and wait for my yes. Open nothing before that.
  7. After a yes, run firstrun pr <id> --json and run the gh commands it prints.
  8. Last, print the line that starts with "share:" from the run output, so I can watch the replay. </pasted_content id="b4ef">
yesterday
f84cfdeRuns and audit pages match the plain landing: no gradient wash, no blur PR 24 gave /runs and /audit the landing's look. The landing has since dropped its gradient wash, glass blur and heavy shadow, so these two pages drop them too. The shared nav, light mode and layout stay. Entire-Checkpoint: 01M4KGD8D11QX44E6CZ2THHB7B+7/-9
590e84aFor Judges page answers the official criteria and the 7 official evidence rows Each of the 5 categories opens with its official criterion and weight. The seven claims block is now the official Evidence and Verification table, in the official order, with what FirstRun gives and a status. New proof on the page: - 4 neatlogs screenshots with the trace or file behind each one - Real git trailer, graph.ndjson and jury self-test output for Entire - A cfo.ai table: tiers, cost per run, caps, revenue, month 18 cash and runway - The recovered trace 5520c5cf as a span sequence - The demo chapters and a timestamped map of each tool's coverage - Commands to list commit dates and count checkpoint trailers - Receipts, Doc-audit Top 10 and team map, and estimated audit costs Entire-Checkpoint: 01M4NCA71P6ST4ATY3FCGSGYM0+293/-73
dc3b781For Judges page: visual proof and half the text The default view drops from about 2,880 words to about 1,550. Each section now shows the official criterion, one visual, at most 3 "Yes" rows and a "Show details" block that keeps every number and link. Visuals, all drawn from repo data: - A larger rubric weights bar - The 7 official evidence rows as a checklist with a glance box and status - v1 against v2 as paired bars (before_after_traces.json) - Entire checkpoint coverage as one stacked bar, 75 of 113 (entire.md) - The cfo.ai tier strip and an inline SVG cash chart (cfo-model.json) - The demo as a timeline bar linked to each chapter second - The 11 tagged posts on a time axis with the 12:30 kickoff marked - A 2-up neatlogs gallery and a few number tiles with source links Entire-Checkpoint: 01M4ND4PYCET73MJBMZQGKF51R+533/-419
891a0c8For Judges page: compact contact sheets for every tool screenshot The page drops from about 10,260 px to about 5,080 px at 1280 wide, and the default view from about 1,550 words to about 790. - A sticky bar keeps the 5 weighted categories one click away - The 7 official evidence rows sit in a two-column grid, one line each - neatlogs, Entire and cfo.ai each get one headline number and a contact sheet: 10, 6 and 5 screenshots, each with a numbered chip and a caption that names its trace or file where the repo records one - Thumbnails open full size in a native dialog: Enter opens, arrows move, Escape closes - New cfo.ai images: the AccuKnox receipt and two crops of the plan page - Entire coverage is recomputed from git on main: 77 of 115 - The 4 upstream PRs show as a screenshot row - Each section has one "Show details" block Entire-Checkpoint: 01M4NE45FC6Q91RYZ6G2WRHBVC+432/-502
37c035cagent.md: fix the friction a fresh agent hit on its first run A sub-agent that had only agent.md audited fastapi/typer and listed 14 friction points. This fixes the ones in the doc and the code. - Windows forms of the preflight and the install (py -3.11, WSL docker, Activate.ps1, PowerShell 7 and UTF-8 redirects). - doctor retries a network probe once, and its fix text says to rerun. agent.md says when to use --for audit. - top_tasks can hold fewer than 10 tasks. The summary keys, ancient, UTC ids, absolute paths and the Jev-off meaning of stale are named. - doctor names the env file it read. Entire-Checkpoint: 01M4NEEKRRHFK8PB8H2F7SNNSZ+98/-28
7fc375bagent.md: give the measured time for a large audit dbt-core with docs.getdbt.com ran for more than 30 minutes on Windows on 2026-10-11, so the 20 minute timeout in agent.md and the skill was wrong. Both now say to run a large audit as a background job or to use the hosted audit. Entire-Checkpoint: 01M4NF3T9C27GDMGNKMRYZEBYQ+3/-3
e11adfbTest count is 384 offline tests after the agent-friendly work Entire-Checkpoint: 01M4NFA6HNR02VBDAV2ZBKGAX6+3/-3
41b9421Agent Ready dialog resets button padding and hover color The homepage styles every button blue with 22 px padding. The dialog now sets padding, border and the hover background on each of its own buttons. Entire-Checkpoint: 01M4NHZQCPCFJEKCDF63GQD5RD+4/-3
f1caf5aFinal alignment: one dated Entire count across the README, evidence and judges page 91 of 129 non-merge commits on main up to b88bcc9 carry an Entire-Checkpoint trailer (38 do not: the 4 setup commits and site commits made while the commit hook hung). 87 checkpoint refs are on GitHub. 76 receipts. 384 offline tests. Entire-Checkpoint: 01M4NJRQ19FYHNXRCSX5Z1NGJZ+10/-10
d8b4ccaAlignment refresh: 91 of 132 commits up to f5d73eb carry a checkpoint, 88 refs Entire-Checkpoint: 01M4NM595AM8KJ27EDET64Q36E+11/-11