From b12386459560b13b0e28a7a854166b59b58d832c Mon Sep 17 00:00:00 2001 From: itsdoubleem <328792799+itsdoubleem@users.noreply.github.com> Date: Mon, 9 Nov 2026 13:24:40 +0900 Subject: [PATCH 1/2] chore: stop naming the assistant notes file in public code (N8) Each comment now names the rule it meant: no invented numbers, check a claim on screen, nothing in src/ names an app, no developer name on app pages. README points at each app's own README as its authority. --- README.md | 2 +- src/components/AppCard.astro | 4 ++-- src/components/OneTrueThing.astro | 4 ++-- src/components/Showcase.astro | 4 ++-- src/content.config.ts | 4 ++-- src/pages/apps/[slug].astro | 4 ++-- src/pages/apps/[slug]/guide.astro | 4 ++-- src/pages/apps/index.astro | 4 ++-- src/pages/index.astro | 6 +++--- src/styles/global.css | 2 +- 10 files changed, 19 insertions(+), 19 deletions(-) diff --git a/README.md b/README.md index c2128e4..ed215a9 100644 --- a/README.md +++ b/README.md @@ -120,7 +120,7 @@ files are committed deliberately, not by accident. | LOGGER | `~/Logger_app` | `public/logger/`, `public/downloads/logger.apk` | | HANGIL | `~/Hangil_app` | `public/hangil/`, `public/downloads/hangil.apk` | -Each has its own `DEPLOY.md` with the exact copy step, and its own `CLAUDE.md` that is the +Each has its own `DEPLOY.md` with the exact copy step, and its own README that is the authority on what that app is. HANGIL's page screenshots are captured from the shipping build by `tools/capture-hangil.mjs` and its guide's by `tools/capture-hangil-guide.mjs`, the same way LOGGER's guide is by `tools/capture-logger.mjs`. diff --git a/src/components/AppCard.astro b/src/components/AppCard.astro index d23edcf..e22e813 100644 --- a/src/components/AppCard.astro +++ b/src/components/AppCard.astro @@ -15,8 +15,8 @@ * do?". It costs one short line. * * And publik's meta line ends `▲ 89% · 3,887 · Download` — a rating, an install count - * and a call to action. None of those exist here and none may be invented: CLAUDE.md - * hard rule 1. What fills the same slot is the facts the frontmatter actually holds. + * and a call to action. None of those exist here and none may be invented: the site's + * rule against invented numbers. What fills the same slot is the facts the frontmatter actually holds. */ import Preview from './Preview.astro'; import { platformShort } from '../labels'; diff --git a/src/components/OneTrueThing.astro b/src/components/OneTrueThing.astro index 5c22682..6c35d0d 100644 --- a/src/components/OneTrueThing.astro +++ b/src/components/OneTrueThing.astro @@ -679,7 +679,7 @@ e rather than shipped: the earlier figure was given as 150,000 years (the accepted benchmark was Omo Kibish at about 195,000; Herto, at about 160,000, is the nearest thing to 150,000) and the Jebel Irhoud paper was given as 2019 (it is - Hublin et al., Nature, June 2017). CLAUDE.md hard rule 1 applies to a paragraph + Hublin et al., Nature, June 2017). The rule against invented numbers applies to a paragraph about honest evidence more than to anywhere else on the site. The date sits in the Jebel Irhoud sentence rather than in an opening "Until 2017" @@ -697,7 +697,7 @@ e ATTEMPT ONE was "Take chapter {primerChapter}, further down." That is simply false about the page, and it was confident about it: it sent the reader hunting - for a card that is not there. CLAUDE.md hard rule 3, in the paragraph about not + for a card that is not there. The check-it-on-screen rule broken, in the paragraph about not overstating evidence, written without once using the UI it described. ATTEMPT TWO made the words a control — a