Skip to content

MANUAL RUN: score the Git Recon menus yourself (#3575) #3576

Open

Description

@QuinnsCode

About 10 minutes. Score the Git Recon menus yourself out of 10, and pick the direction for the next round.

The menus climb (#3575, target 8/10) stalled tonight at about 5 on agent judges' scores. Agent judges run about two points high on these climbs and drift between judges, so your score is the real one.

Steps

  1. bun run manual and pick this issue: it opens each screen below. Or by hand: cd apps/git-recon && bun run dev, then open these paths on the port it prints:

    • /: the title page and the missions as dog tags
    • /?screen=loadout: the build, head and role pickers, as a parts tray
    • /?screen=settings: the Controls panel
    • /?screen=won and /?screen=lost: results, with a sample round's numbers
    • Each again at phone width (devtools, 390 px).

    The ?screen= links come with PR Git Recon: baked menu art, a sprue of clay soldiers behind the page, the soldier on results (menus climb round 4) #3574. Until it merges, open the title and click: "More" for loadout, the gear for settings. Win or lose Boot camp for results.

  2. Score the menus out of 10 on the goalpost in Git Recon: menus as a toy-soldier briefing, hill climb to 8/10, rounds 5+ (from #3569) #3575:

  3. Which direction for the next round?

    • More toy box: sprues, clay and moulded parts everywhere. For example, the loadout as a real tray of parts you pick, and the figures resculpted for the renders.
    • Back toward the field manual: paper, stencils and diagrams, with the toy box as an accent.
    • Something else.

    Comment with your score and the direction. Yes closes this; No (with a note) keeps it open.

Before and after (desktop)

main, before round 1 now (round 4, #3574)
title
loadout
settings
results

Every round at desktop, tablet and phone is in apps/git-recon/docs/shots-2026-11-04-menus/ (before-*, r1-* to r3-*, now-*).

The judges' numbers

Each round had a fresh judge, scored out of 10 and corrected two points down:

round judge's BEFORE judge's NOW
1 1.5 5.0
2 4.4 5.2
3 4.8 5.3
4 4.5 5.2

The drift between judges is as big as the gains: the stall note on #3575 has the detail.

For agents

  • bun run probe git-recon "/?screen=won" --shot out.png shows any screen. In dev builds, window.__recon has the game.
  • bun scripts/menu-art.ts in apps/git-recon re-bakes the backplate and the result soldiers.

Activity

  1. added
    manual runNeeds the owner to run it by hand (tokens, sign-ins); agents can't close it
    on Nov 5, 2026
  2. QuinnsCode commented on Nov 5, 2026

    @QuinnsCode
    OwnerAuthor

    The owner's direction, given in chat to Nibble on 2026-11-05: move away from the olive toy box toward a sniper or spy-thriller mission briefing (scope reticle, dossier, classified marks) on the clean white crafting-sandbox canvas, with chip tiles and one hot accent, but rugged. Keep the dog tags. Round 5 builds that: PR #3593 (shots in apps/git-recon/docs/shots-2026-11-05-briefing/). The climb continues in #3594. This issue stays open for your own score.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    manual runNeeds the owner to run it by hand (tokens, sign-ins); agents can't close it

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions