@tideworthy/tidysites 0.1.24 → 0.1.25

package/README.md

@@ -96,6 +96,10 @@ names its own in `header_row`.
 5. The look: brand colours and fonts on the Business record, `business/business.css` for what the tokens cannot say. The theme's own rules are written as `.ts.ts-accent` and `.ts.ts-ink .ts-prose a`, so a rule of yours needs the same weight (`.ts.ts-accent …`) to win.
 6. `tidysites validate`, `tidysites push`, `tidysites shot / --all`, and look at every width: a design often shows a different header above 1440px.
 
+## The website's own instructions
+
+A website can carry standing instructions from the partner on how it is done. `pull` (and `init`) write them into `AGENTS.md` in the folder, with a `CLAUDE.md` that imports it (`@AGENTS.md`), and keep their version in `.tidysites/snapshot.json`. Both land as a marked block (`<!-- tidysites:instructions -->`), so a file you already keep under either name stays yours: the block is appended to it, replaced in place on every pull, and taken out when the website has no instructions. Read them before changing anything. `push`, `redirect`, and `publish` send that version with the change; when the website's instructions changed since the last pull, they stop and say to pull again. A website with no instructions gets no files, and ours are taken away on the pull that finds them gone (CLAUDE.md only while it is still the one line we wrote).
+
 ## The skill
 
 `skill/SKILL.md` is the Working with Tidysites skill: how a site is put together, where every setting lives, the order to port or build, and this command line's part in it. It is the one copy. The connector serves it as `get_tidysite_skill`, the platform app imports it from this package, and `tidysites skill` writes it into a project. A change to a command here, to the content model, or to where a setting lives on the template edits `skill/SKILL.md` in the same pull request, and ships with the next `cli-v*` tag.

package/bin/tidysites.mjs

@@ -23,11 +23,16 @@
  * The site is the remote: pull, edit, push. A push that would overwrite an
  * entry changed on the website since the pull is refused until you pull
  * again (or say --force).
+ *
+ * A website's own instructions (the partner's standing word on how the
+ * site is done) come down with the pull as AGENTS.md, with a CLAUDE.md that
+ * imports it. Read them first. Every change sends their version back, and
+ * a push is refused when they changed since the last pull.
  */
 import { existsSync, mkdirSync, readdirSync, readFileSync, rmSync, statSync, writeFileSync } from 'node:fs';
 import { dirname, join, relative, resolve } from 'node:path';
 import { callTool, ApiError } from '../lib/api.mjs';
-import { diffEntries, entryToFiles, filesToEntry, FILE_NAME, menuItems, menuTree, pushDocument, sameEntry, snapshotOf, SIDECARS, splitDocument, documentWeight, mergePushResults } from '../lib/folder.mjs';
+import { diffEntries, entryToFiles, filesToEntry, FILE_NAME, menuItems, menuTree, pushDocument, sameEntry, snapshotOf, SIDECARS, splitDocument, documentWeight, mergePushResults, withBlock, CLAUDE_LINE } from '../lib/folder.mjs';
 import { validateAll, indexSchema, validateMenuNames } from '../lib/validate.mjs';
 import { cleanHtml, cssSummary, fetchPage, outline, outlineText } from '../lib/fetch.mjs';
 
@@ -184,9 +189,10 @@ async function pull({ dir, cfg }, collections = null, { overwrite = new Set(), m
 		await pullMenus({ dir, cfg });
 		return;
 	}
-	const [{ json: doc }, { json: schema }] = await Promise.all([
+	const [{ json: doc }, { json: schema }, { json: site }] = await Promise.all([
 		tool(cfg, 'pull_website', { site_id: cfg.site, ...(want ? { collections: want.join(',') } : {}) }),
-		tool(cfg, 'get_website_schema', { site_id: cfg.site })
+		tool(cfg, 'get_website_schema', { site_id: cfg.site }),
+		tool(cfg, 'get_website', { site_id: cfg.site })
 	]);
 	if (!doc?.content) fail('The pull came back without content.');
 	const pulled = new Set(doc.collections.map((c) => c.slug));
@@ -248,6 +254,10 @@ async function pull({ dir, cfg }, collections = null, { overwrite = new Set(), m
 		snap.settings = prev.settings;
 		snap.collections = [...new Set([...(prev.collections ?? []), ...snap.collections])];
 	}
+	// The website's own instructions: the words as AGENTS.md, their version in the snapshot for every change to carry.
+	const instructions = site?.website?.instructions ?? null;
+	snap.instructions_version = site?.website?.instructions_version ?? null;
+	const hadInstructions = writeInstructions(dir, instructions);
 	writeFileSync(join(dir, META, 'snapshot.json'), JSON.stringify(snap, null, '\t'));
 	writeFileSync(join(dir, META, 'schema.json'), JSON.stringify(schema, null, '\t'));
 	const n = doc.collections.reduce((a, c) => a + c.count, 0);
@@ -255,6 +265,7 @@ async function pull({ dir, cfg }, collections = null, { overwrite = new Set(), m
 	for (const c of doc.collections.filter((c) => c.count)) say(`  ${c.slug.padEnd(20)} ${c.count}`);
 	if (snap.menus) say(`  ${'menus'.padEnd(20)} ${Object.keys(snap.menus).length}`);
 	for (const k of kept) say(`  kept your unpushed changes in ${k}`);
+	if (hadInstructions) say(`  AGENTS.md: the website's own instructions${prev && prev.instructions_version !== snap.instructions_version ? ' (changed since the last pull)' : ''}; read them before changing anything`);
 	if (heldBack.length) say(`  ${heldBack.length} entr${heldBack.length === 1 ? 'y' : 'ies'} you removed here stay removed (the website still has them until \`push --prune\`)`);
 }
 
@@ -326,6 +337,7 @@ async function push(ctx) {
 		say('Nothing to push.');
 		return;
 	}
+	await requireCurrentInstructions(ctx);
 	const doc = pushDocument(s.local, s.snap, { keys, removeKeys });
 	if (flags.force) doc.force = true;
 	// A picture that exists only as a file here (an illustration drawn from the old page's inline SVG, a rasterised
@@ -346,7 +358,7 @@ async function push(ctx) {
 			const parts = splitDocument(doc);
 			if (parts.length > 1) say(`${parts.length} parts: the folder's pictures and videos weigh ${(documentWeight(doc) / 1048576).toFixed(0)} MB.`);
 			for (const [i, part] of parts.entries()) {
-				const { json } = await callTool(ctx.cfg, 'push_website', { site_id: ctx.cfg.site, ...part });
+				const { json } = await callTool(ctx.cfg, 'push_website', { site_id: ctx.cfg.site, ...stamp(ctx), ...part });
 				if (!json || typeof json !== 'object' || !json.content) throw new ApiError(`The website did not answer the push${parts.length > 1 ? ` (part ${i + 1} of ${parts.length})` : ''} in time or in a readable way`, { status: 504 });
 				r = r ? mergePushResults(r, json) : json;
 			}
@@ -376,13 +388,13 @@ async function push(ctx) {
 		}
 		try {
 			try {
-				await tool(ctx.cfg, 'call_website_site_tool', { site_id: ctx.cfg.site, tool: 'menu_set_items', arguments: { name, items } }, { throws: true });
+				await tool(ctx.cfg, 'call_website_site_tool', { site_id: ctx.cfg.site, tool: 'menu_set_items', arguments: { name, items }, ...stamp(ctx) }, { throws: true });
 			} catch (e) {
 				// A menu the website does not have yet (a phone drawer, a footer column) is created from the file, then filled.
 				if (!/not.?found/i.test(e.message ?? '')) throw e;
 				if (!/^[a-z][a-z0-9_]*$/.test(name)) throw new Error(`a new menu's name is lowercase letters, digits and underscores (${name} has something else)`);
-				await tool(ctx.cfg, 'call_website_site_tool', { site_id: ctx.cfg.site, tool: 'menu_create', arguments: { name, label: m.label || name } }, { throws: true });
-				await tool(ctx.cfg, 'call_website_site_tool', { site_id: ctx.cfg.site, tool: 'menu_set_items', arguments: { name, items } }, { throws: true });
+				await tool(ctx.cfg, 'call_website_site_tool', { site_id: ctx.cfg.site, tool: 'menu_create', arguments: { name, label: m.label || name }, ...stamp(ctx) }, { throws: true });
+				await tool(ctx.cfg, 'call_website_site_tool', { site_id: ctx.cfg.site, tool: 'menu_set_items', arguments: { name, items }, ...stamp(ctx) }, { throws: true });
 				say(`Menu ${name}: created.`);
 			}
 			say(`Menu ${name}: ${items.length} item${items.length === 1 ? '' : 's'} written.`);
@@ -533,7 +545,8 @@ async function redirect(ctx) {
 	if (sub === 'add') {
 		const [, from, to] = positional;
 		if (!from || !to) fail('Usage: tidysites redirect add <old path> <new path or url>');
-		const { json } = await tool(ctx.cfg, 'add_website_redirect', { site_id: ctx.cfg.site, source: from, destination: to });
+		await requireCurrentInstructions(ctx);
+		const { json } = await tool(ctx.cfg, 'add_website_redirect', { site_id: ctx.cfg.site, source: from, destination: to, ...stamp(ctx) });
 		say(json ? `${from} → ${to}` : 'Added.');
 		return;
 	}
@@ -544,7 +557,8 @@ async function redirect(ctx) {
 		const rows = list?.redirects ?? list?.items ?? [];
 		const hit = rows.find((r) => (r.source ?? r.from) === from || (r.source ?? r.from) === from.replace(/\/$/, '') || (r.source ?? r.from) === `${from}/`);
 		if (!hit) fail(`No redirect from ${from}.`);
-		await tool(ctx.cfg, 'remove_website_redirect', { site_id: ctx.cfg.site, redirect_id: hit.id });
+		await requireCurrentInstructions(ctx);
+		await tool(ctx.cfg, 'remove_website_redirect', { site_id: ctx.cfg.site, redirect_id: hit.id, ...stamp(ctx) });
 		say(`Removed ${from}.`);
 		return;
 	}
@@ -683,10 +697,42 @@ function readRemoved(dir) { return new Set(readJson(join(dir, META, REMOVED_FILE
 function writeRemoved(dir, set) { writeFileSync(join(dir, META, REMOVED_FILE), JSON.stringify([...set].sort(), null, '\t')); }
 function rememberRemoved(dir, key) { const s = readRemoved(dir); s.add(key); writeRemoved(dir, s); }
 
+/**
+ * The website's own instructions as files. AGENTS.md is the file every assistant working in a folder opens
+ * first, and CLAUDE.md is the one Claude Code reads; a person may already keep either. So the words live in a
+ * marked block: the whole file when there was none, appended to theirs when there was, replaced in place on
+ * every pull, and taken out again (the file too, when nothing else is in it) once the website has none.
+ */
+function writeManaged(path, body) {
+	const existing = existsSync(path) ? readFileSync(path, 'utf8') : null;
+	const next = withBlock(existing, body);
+	if (next === null) { if (existing !== null) rmSync(path); return; }
+	if (next !== existing) writeFileSync(path, next);
+}
+function writeInstructions(dir, text) {
+	const has = Boolean(text && text.trim());
+	writeManaged(join(dir, 'AGENTS.md'), has ? text.replace(/\s+$/, '') : null);
+	writeManaged(join(dir, 'CLAUDE.md'), has ? CLAUDE_LINE : null);
+	return has;
+}
+/** The version of the instructions the folder last pulled, sent with every change as proof they were read. */
+function stamp({ dir }) {
+	const v = readJson(join(dir, META, 'snapshot.json'))?.instructions_version;
+	return v ? { instructions_version: v } : {};
+}
+/** Before a change: the website's instructions must be the ones this folder pulled. */
+async function requireCurrentInstructions(ctx) {
+	const { json } = await tool(ctx.cfg, 'get_website', { site_id: ctx.cfg.site });
+	const theirs = json?.website?.instructions_version ?? null;
+	const mine = readJson(join(ctx.dir, META, 'snapshot.json'))?.instructions_version ?? null;
+	if (theirs !== mine) fail("The website's instructions changed since the last pull. Run `tidysites pull`, read AGENTS.md, then try again.");
+}
+
 async function publish(ctx) {
 	if (!flags.yes) fail('Publishing makes the live site a copy of the preview site. Look at it first (`tidysites shot /`), then run `tidysites publish --yes`.');
 	// A website live on the client's own domain is never published from here: the website refuses (409 live_domain) and says who may.
-	const { json, text } = await tool(ctx.cfg, 'publish_website', { site_id: ctx.cfg.site });
+	await requireCurrentInstructions(ctx);
+	const { json, text } = await tool(ctx.cfg, 'publish_website', { site_id: ctx.cfg.site, ...stamp(ctx) });
 	say(json ? JSON.stringify(json, null, 2) : text);
 }
 
@@ -722,6 +768,7 @@ async function main() {
 			writeFileSync(join(dir, META, 'key'), `${key}\n`, { mode: 0o600 });
 			writeFileSync(join(dir, META, '.gitignore'), 'key\nshots/\nsource/\npages/\n');
 			say(`Connected ${relative(process.cwd(), dir) || '.'} to ${site} on ${host}.`);
+			// The pull below writes AGENTS.md when the website carries its own instructions, and says so.
 			// A website that is still being set up answers 404 on its content model for a few minutes; the first pull waits for it.
 			for (let attempt = 1; ; attempt++) {
 				try { await pull(config()); break; }

package/lib/folder.mjs

@@ -264,3 +264,21 @@ export function splitDocument(doc) {
 export function mergePushResults(a, b) {
 	return { ...a, ...b, content: { created: (a.content?.created ?? 0) + (b.content?.created ?? 0), updated: (a.content?.updated ?? 0) + (b.content?.updated ?? 0) }, images: (a.images ?? 0) + (b.images ?? 0), images_reused: (a.images_reused ?? 0) + (b.images_reused ?? 0), missing: [...(a.missing ?? []), ...(b.missing ?? [])] };
 }
+
+/**
+ * The website's own instructions inside a file a person may already keep (AGENTS.md, CLAUDE.md): a marked
+ * block, the whole file when there was none, appended to theirs when there was, replaced in place on every
+ * pull. `body` null takes the block out; the result is null when nothing else was in the file.
+ */
+export const BLOCK_START = "<!-- tidysites:instructions (the website's own instructions; rewritten on every pull, edit them in the app) -->";
+export const BLOCK_END = '<!-- /tidysites:instructions -->';
+/** The one line Claude Code needs in CLAUDE.md to read AGENTS.md. */
+export const CLAUDE_LINE = '@AGENTS.md';
+const escapeRe = (s) => s.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
+const BLOCK_RE = new RegExp(`\\n*${escapeRe(BLOCK_START)}[\\s\\S]*?${escapeRe(BLOCK_END)}\\n*`);
+export function withBlock(existing, body) {
+	const rest = (existing ?? '').replace(BLOCK_RE, '\n').trim();
+	if (body === null) return rest ? `${rest}\n` : null;
+	const block = `${BLOCK_START}\n${body.trim()}\n${BLOCK_END}`;
+	return rest ? `${rest}\n\n${block}\n` : `${block}\n`;
+}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tideworthy/tidysites",
-  "version": "0.1.24",
+  "version": "0.1.25",
   "description": "Pull a Tidysites website to files, edit, validate against its content model, push it back, screenshot the result.",
   "type": "module",
   "bin": {

package/skill/SKILL.md

@@ -11,6 +11,10 @@ A Tidysite is a template (the data shape every site shares), a theme (the look),
 
 Content work happens on the preview site. The live site changes only when `publish_website` runs or a person presses Publish now.
 
+## The website's own instructions
+
+A website can carry standing instructions from the partner: what never changes, how the copy reads, which pages are theirs to keep. They say how this site is done; this document says how Tidysites works. In a folder connected with the command line they are `AGENTS.md` (CLAUDE.md imports it), pulled with the content; read them before the first change. Every change the command line makes sends their version, and a push is refused when they changed since the last pull (pull again, read AGENTS.md). Through the connector, `get_website` returns them as `instructions` with an `instructions_version`, and every tool that changes the site takes that version back as `instructions_version`.
+
 ## The three layers
 
 1. **Records.** The things with a page of their own or reused across pages: the Business record, Locations, Services and Service categories, Service areas, Audiences (who a service is for), People, Posts, Products, Case studies, Pages, Forms, Custom rows, Saved sections, and the FAQs and Reviews when they are shared. Logos, insurers, financing providers, technologies, conditions, policies, downloads, offers and badges are written in their section as items; a collection of them exists only for the shared case. Read and write them with `list_website_content`, `get_website_entry`, `create_website_entry`, `update_website_entry`, `edit_website_section` and `replace_website_text`.