Cuter sprites

[dvd90]

The v0.1 creature read as boxy — this redesigns it to actually be cute (the pet is the UI).

Rounded body, wide-set expressive eyes per mood, little antennae + feet on the kid, a footless chibi baby, a speckled egg. Drawn with single-width box-drawing glyphs, so the terminal width math (line.length === display width) is unchanged and the ≤12×6 invariant still holds.

• src/ui/sprites/byte.ts redesigned
• Snapshots regenerated (sprites, App, card)
• Repo share card regenerated with the new art

Preview of all moods + growth stages: rendered in the PR discussion / shared separately.

🤖 Generated with Claude Code

[claude]

Claude finished @dvd90's task in 2m 53s —— View job

---

Review: Cuter sprites (PR #5)

• Gather context (diff, CLAUDE.md conventions)
• Review src/ui/sprites/byte.ts changes (sprites-as-data, ≤12×6 invariant, display-width math)
• Review snapshot updates (sprites, App, card)
• Check TDD conventions (tests updated alongside sprite changes)
• Post final review

Scope check: only src/ui/sprites/byte.ts, its snapshots, and the regenerated share-card SVG changed. No src/engine/** or src/collectors/** touched, so purity/never-throw rules aren't implicated here. Sprites stay as plain TS data objects (not inlined in components) — consistent with the "sprites are data" rule. No new deps, named exports only, commit is a proper feat:.

Main finding — the "display width == .length" claim needs verifying, not assuming

byte.ts:5-6 asserts the new glyphs are safe because "box-drawing and geometric glyphs are width 1." The box-drawing characters (╭╮┴│╰┬╵) are indeed universally narrow, but a few of the new eye/mouth glyphs live in Unicode ranges marked Ambiguous in EastAsianWidth.txt — notably • (U+2022, General Punctuation) and parts of the Geometric Shapes / Miscellaneous Technical blocks that ◕ ◔ ⌣ are drawn from. "Ambiguous" characters render as 1 column in most Western-locale terminals but as 2 columns in many CJK-locale terminals and in apps configured with ambiguous-wide (a common tmux/iTerm2/Windows Terminal setting).

The existing invariant test (test/ui/sprites.test.ts:17-19) only checks frame.split('\n') line .length equality — that's UTF-16 code-unit count, not terminal display width, and every one of these glyphs is a single code unit, so the test passes regardless of how it actually renders. That means the exact scenario the PR body calls out ("terminal width math unchanged") isn't actually verified — it's asserted in a comment and happens to pass a test that can't distinguish it from the real risk. If any of these glyphs render 2-wide in a user's terminal, rows in the same frame padded with .length would visually misalign, breaking the ≤12×6 boxed look the redesign is going for — silently, and only for a subset of terminal configs, which makes it easy to miss in review/CI.

Suggested options (not necessarily blocking, given this is a low-severity visual-only risk and the ambiguous-width terminal configs are a minority case):

• Swap •, ◕, ◔, ⌣ for glyphs confirmed Narrow/Neutral (many box-drawing/block-element alternates exist), or
• Add a dev-time check using the same width logic Ink itself uses internally (string-width) against these specific glyphs, so a future glyph swap can't silently reintroduce this.

Fix this →

Everything else

• frame() padding math checked by hand for kid (8×6), baby (6×5), egg (6×5) — all rows within a given frame are equal width and ≤12×6, matching the existing invariant test.
• Egg's two frames (speckleRight true/false) are same width, so the shimmer animation won't jitter.
• Snapshots (sprites, app, card) are consistent with the new glyphs and regenerated correctly — no stray diffs.
• This is a data-only redesign already covered by the existing generalized invariant test (test/ui/sprites.test.ts:9), so no new tests were strictly required per TDD — the meta-test already exercises every (stage, mood) combo against the size/padding rules.
• Note: I wasn't able to run npm test/typecheck/lint myself (no node_modules, and installing requires elevated approval not available here) — the hand-checked math above stands in for that, but please confirm CI is green before merge.

Overall: nice visual upgrade, keeps within the architectural rules. The one thing worth a second look before merging is the ambiguous-width glyph risk above — everything else looks good.

[claude]

Claude finished @dvd90's task in 1m 18s —— View job

---

Review: Cuter sprites (PR #5) — update

• Gather context (diff vs origin/main, CLAUDE.md conventions)
• Review what changed in the new commit (4f3ae34)
• Check TDD, engine purity, collectors-never-throw, product principles
• Post final review

What's new since the last review: commit 4f3ae34 is docs-only — CONTRIBUTING.md, docs/ARCHITECTURE.md, docs/manual-tests.md, README rewrite, docs/demo.gif/demo.tape, and a regenerated docs/gitgotchi-card.svg (egg art). src/ui/sprites/byte.ts itself is byte-for-byte the same as it was for the first review — no code changed in this push.

Docs spot-check: the new architecture/README claims were cross-checked against the actual source and hold up — e.g. the life-stage table (README.md: baby @ 3 check-ins, kid @ 20 check-ins + coverage ≥ 50%, adult @ 100 + green streak ≥ 10, elder @ 365 days) matches src/engine/evolution.ts:19-22 exactly. ARCHITECTURE.md's data-flow/layer description matches the real module layout. No src/engine/** or src/collectors/** code was touched, so purity/never-throw rules aren't implicated by this push either — the new docs just describe those rules accurately, they don't loosen them.

Previous finding still open, unaddressed by this push: src/ui/sprites/byte.ts:8-24 uses •, ◕, ◔, ⌣, which fall in Unicode's Ambiguous East Asian Width category — they render as 1 column in most Western-locale terminals but 2 columns under CJK locales or ambiguous-wide terminal settings (common in tmux/iTerm2/Windows Terminal). The invariant test (test/ui/sprites.test.ts:9-22) only checks .length (UTF-16 code units), which can't detect this, so the ≤12×6/equal-width guarantee the PR body claims isn't actually verified for the risky glyphs. This is low-severity and cosmetic-only (a minority of terminal configs), so not necessarily blocking, but worth a decision before merge rather than silently shipping:

• swap the four glyphs for confirmed-Narrow alternatives, or
• add a one-time dev check with string-width (already a transitive dep via Ink) asserting each glyph in EYES/MOUTH/BLINK is width 1.

Fix this →

Everything else in this push: the docs additions are well-scoped (contributing guide, architecture tour, manual-test playbook, README overhaul with a demo GIF) — none of it touches src/engine/** or src/collectors/**, so no new TDD/purity obligations are introduced. docs/gitgotchi-card.svg regeneration matches the new egg sprite. No new runtime deps added. Conventional commit messages used throughout.

Note: I still can't run npm test/typecheck/lint myself in this sandbox (no node_modules, no install permission) — please confirm CI is green before merge.

Overall: nice docs pass, no regressions introduced. The one open item is the ambiguous-width glyph risk from the original review, carried over unchanged.