Presets

A preset holds everything that differs between kinds of web app: the settings it needs, how you sign in, CSS that hides what changes from day to day, the checks and steps a brief can name, and where release-check finds the latest version. The capture engine itself knows no product. Presets live in scripts/presets/<name>.json; a config picks one with "preset", and a brief lists the ones it uses in "presets".

Preset	Sign-in	Release source	Starter brief
p1-editor	You, in the dedicated Chrome	npm:@pantheon-systems/p1-next-sdk	briefs/p1-editor.json
wordpress-admin	Login form: WP_USER, WP_PASSWORD	wordpress	briefs/wordpress-admin.json
drupal-admin	Login form: DRUPAL_USER, DRUPAL_PASSWORD	drupal	briefs/drupal-admin.json
content-publisher	You, in the dedicated Chrome	(none; pass --version or --source)	briefs/content-publisher.json
pantheon-dashboard	You, in the dedicated Chrome	release notes tagged User interface (watch)	briefs/pantheon-dashboard.json
gdocs-addon	You, in the dedicated Chrome	release notes tagged Content Publisher (watch)	briefs/gdocs-addon.json
word-addin	You, in the dedicated Chrome	release notes tagged Content Publisher (watch)	briefs/word-addin.json
public-site	None	(none; pass --version or --source)	briefs/public-site.json

p1-editor

Actions for the P1 editor (the /p1 route by default). Selectors come from the editor’s data-testid attributes and from Puck’s data-puck-component attribute on canvas blocks, which holds the same props.id the P1 MCP returns from get_document. Values such as the project, page, workstream, and block type are parameters: the brief passes them in, and they come from the per-run config file. Re-check the selectors after a P1 editor upgrade. editorReady also restores the left panel, because the editor keeps panel state across page loads in the same browser. The block-selection actions end with a selectedBlock check: exactly one selected overlay, matching the intended block’s box, and the properties panel naming its type. A neighboring block fails it.

Settings (under params in the config, or at the top level in configs written before presets):

Key	Required	Default	Meaning
projectName	yes		The project or site name as the editor header shows it. Find it with the P1 MCP (list_sites) or in the editor.
workstream	yes		The workstream to show, as the editor’s workstream selector lists it. Choose it for each run (P1 MCP list_branches); there is no default.
pagePath	yes		The page being edited, for example “/” or “/about” (P1 MCP list_documents); it must start with “/” and contain no spaces.
pageNavigation	no	url	“url” opens the page through the editor URL; “manual” opens the editor route only and you select the page. One of url, manual.
pageLabel	no	{pagePath}	The page label the editor’s page selector shows (default: pagePath).
editorRoute	no	/p1	The editor route; it must start with “/”.
blockCategory	no	P1 Layout	The block category to expand in the blocks browser.
blockType	no		A block type that selects reliably.
blockSelector	no		A CSS selector for the block to select, instead of blockType.
publishMenuLabel	no	Publish this page to	The publish menu’s label.

Actions:

Action	Parameters	Checks
editorReady	projectName, pageLabel	[data-testid=p1-editor-header]; [data-testid=site-label] contains “{projectName}”; [data-testid=page-selector] contains “{pageLabel}”; [data-testid=workstream-trigger] …
selectWorkstream	name	[data-testid=workstream-trigger] contains “{name}”; h1
openWorkstreamMenu		[data-testid=workstream-dropdown]; [data-testid=workstream-list]
closeMenu
expandAllBlockCategories
expandBlockCategory	name
selectBlockByType	type	selected block
selectBlock	id	selected block
centerBlockByType	type
workstreamMenuClosed		[data-testid=workstream-dropdown]
collapseLeftPanel		[aria-label="Toggle left panel"][aria-pressed=false]; #p1-cat-btn-p1Layout
openPublishMenu	label	[role=menu] contains “{label}”
selectBlockBySelector	selector	selected block
centerBlockBySelector	selector

wordpress-admin

WordPress admin (wp-admin), the block editor, and the Site Editor. Signs in at /wp-login.php with WP_USER and WP_PASSWORD from the environment, for a local or test site; for a production site with SSO or two-factor, set “signIn”: “chrome” in the config and sign in yourself in the dedicated Chrome. Hides admin notices, update counts, and the editor’s welcome guide, so shots don’t depend on what a plugin printed that day. Adapted from the wp-screenshots skill. Selectors checked against WordPress 6.8 on 2026-10-08.

Settings: none beyond topic and baseUrl.

Sign-in: opens /wp-login.php, types the values of WP_USER and WP_PASSWORD from your shell environment, submits #loginform, and expects a URL containing /wp-admin. Set "signIn": "chrome" in the config to sign in yourself instead.

Actions:

Action	Parameters	Checks
adminReady		#wpadminbar; #adminmenu; #wpbody-content
adminPage	title	#wpadminbar; #adminmenu; .wrap h1 contains “{title}”
blockEditorReady		.editor-editor-interface; .editor-header; iframe[name="editor-canvas"]
siteEditorReady		.edit-site-layout; iframe[name="editor-canvas"]

drupal-admin

Drupal administration pages (Claro theme, with the Navigation sidebar in Drupal 11.3 and later or the classic toolbar before it). Signs in at /user/login with DRUPAL_USER and DRUPAL_PASSWORD from the environment, for a local or test site; for a hosted site with SSO, set “signIn”: “chrome” in the config and sign in yourself in the dedicated Chrome. Hides status messages so a leftover “saved” message doesn’t end up in a shot. Selectors checked against Drupal 11.4 on 2026-10-08.

Settings: none beyond topic and baseUrl.

Sign-in: opens /user/login, types the values of DRUPAL_USER and DRUPAL_PASSWORD from your shell environment, submits #user-login-form, and expects a URL containing /user/. Set "signIn": "chrome" in the config to sign in yourself instead.

Actions:

Action	Parameters	Checks
adminReady		#admin-toolbar, #toolbar-administration; main; h1.page-title
adminPage	title	#admin-toolbar, #toolbar-administration; h1.page-title contains “{title}”

content-publisher

Pantheon Content Publisher’s dashboard (content.pantheon.io): overview, collections, sites, and one collection’s settings. You sign in yourself, with Google, in the dedicated Chrome. Hides the support chat launcher, toast notifications, first-visit onboarding dialogs (without dismissing them, which would save a preference to the account), and the buttons some browser extensions inject; a capture profile with no extensions is cleaner still. Selectors checked against the live dashboard on 2026-10-08. The starter brief leaves out Accounts and Tokens, which show account data. A trial account also shows a free-trial banner across the top, which has no stable selector and is left in.

Settings (under params in the config):

Key	Required	Default	Meaning
collectionId	no		A collection ID from the collection’s URL (content.pantheon.io/dashboard/collections/). Shots of one collection's settings need it.

Actions:

Action	Parameters	Checks
dashboardReady		nav[aria-label="Main navigation"]; main#app-layout-main
pageHeading	text	nav[aria-label="Main navigation"]; main h1 contains “{text}”
collectionReady		nav[aria-label="Main navigation"]; nav[aria-label="Collection settings secondary navigation"]; main h1

gdocs-addon

A Google Docs add-on in the Docs side panel, for example Pantheon Content Publisher. You sign in to Google yourself in the dedicated Chrome. openAddon presses the add-on’s button in the right-hand side panel (skipped when it’s already open), waits for the panel, and checks the panel’s text; it clicks nothing inside the add-on, so nothing is connected or published. The panel closes with the tab, so the doc isn’t left with it open. Use a test doc: shots show the doc’s content and your account’s avatar. addonClick clicks a button inside the panel by its text and checks the panel’s text; addonWindow does the same for a button that opens a window (Preview and publish) and captures that window, which is closed afterward. Neither clicks Publish, Save, or Disconnect. Selectors checked against Google Docs on 2026-10-09, where Content Publisher’s button is labeled “Pantheon”.

Settings (under params in the config):

Key	Required	Default	Meaning
documentId	yes		The test doc’s ID, from docs.google.com/document/d//edit.
addonLabel	no	Pantheon	The add-on’s button label in the side panel (its tooltip). Content Publisher’s is “Pantheon”.

Actions:

Action	Parameters	What it does
docReady		#docs-editor; #docs-menubar
openAddon	label, text	click .app-switcher-button[aria-label="{label}"]; check .app-switcher-button[aria-label="{label}"][aria-selected=true]; check “{text}” in the panel
addonClick	text, check	click [role=button], button with “{text}” in the panel; check “{check}” in the panel
addonWindow	text, url, check, busy	click [role=button], button with “{text}” in the panel, follow the window with “{url}”; check “{check}”; check no “{busy}”

word-addin

An Office add-in’s task pane in Word for the web, for example Pantheon Content Publisher. You sign in to Microsoft yourself in the dedicated Chrome, and the add-in must already be added to Word (adding it accepts terms, so it’s yours to do). openAddin clicks the add-in’s ribbon button on the Home tab and waits for the task pane’s text; addinClick clicks a button inside the task pane by its text. The task pane is an iframe inside Word’s editor iframe, so checks use a frame list. Word’s own panels (Navigation, the Copilot drafting prompt on an empty document) are inside Word’s frame and aren’t hidden: close the Navigation pane and put some text in the test document. Selectors checked against Word for the web on 2026-10-09.

Settings (under params in the config):

Key	Required	Default	Meaning
documentUrl	yes		The test document’s full URL from the address bar of Word for the web.
addinName	no	Pantheon Content Publisher	The add-in’s name as its ribbon button and task pane title show it.

Actions:

Action	Parameters	What it does
wordReady		iframe#WacFrame_Word_Inline; in Word: #InsertAddInFlyout
openAddin	name, text	click button[aria-label="{name}"] in Word; check “{text}” in the task pane
addinClick	name, text, check	click button, [role=button], a with “{text}” in the task pane; check “{check}” in the task pane

pantheon-dashboard

The Pantheon dashboard (dashboard.pantheon.io): workspace pages and a site’s environment tabs. You sign in yourself, through Pantheon’s single sign-on, in the dedicated Chrome. Use a test workspace: shots show workspace, site, and team names. Hides the maintenance and warning banners, toast notifications, and the support chat, which change from day to day. Selectors checked against the live dashboard on 2026-10-09; personal settings are /workspace//personal-settings/; site routes are /workspace//cms-site//environment//.

Settings (under params in the config):

Key	Required	Default	Meaning
workspaceId	yes		The workspace ID from the dashboard URL (dashboard.pantheon.io/workspace//home).
siteId	no		A site ID from a site’s URL (…/cms-site//…). Shots of a site's environments need it; use a site that isn't frozen.
env	no	dev	The environment for site shots: dev, test, live, or a Multidev name.

Actions:

Action	Parameters	Checks
dashboardReady		nav[aria-label="Global Primary Navigation"]; main#app-layout-main
workspacePage	title	nav[aria-label="Global Primary Navigation"]; main h1 contains “{title}”
siteTab	heading	nav[aria-label="CMS Site Top Navigation"]; nav[aria-label="CMS Site Environments Navigation"]; main h2 contains “{heading}”
siteReady		nav[aria-label="CMS Site Top Navigation"]; nav[aria-label="CMS Site Environments Navigation"]; main h2
personalPage	heading	nav[aria-label="Global Primary Navigation"]; main h1 contains “User settings”; main h2 contains “{heading}”

public-site

Any public website: no sign-in. Hides the cookie and consent banners of common consent platforms (OneTrust, Cookiebot, CookieYes, Osano, Usercentrics, Cookie Notice, Cookie Consent), so shots don’t depend on a stored consent choice, and the chat widgets of Qualified, Intercom, HubSpot, Drift, Zendesk, and Crisp. Qualified’s docked panel also narrows the page with a margin on <html>; the preset removes it, but only on pages that have that panel. A site behind a bot challenge fails closed: no PNG is saved.

Settings: none beyond topic and baseUrl.

Actions:

Action	Parameters	Checks
visible	selector	{selector}
heading	text	h1 contains “{text}”

Write a preset

Copy the closest preset and change what differs. Every field is optional except description and actions:

Field	Meaning
description	What the preset is for, which version its selectors were checked against, and when.
params	Settings the preset needs from the config: { "name": { "required", "what", "pattern", "enum", "default" } }. A default can name another setting, as in "{pagePath}".
atMostOne	Groups of settings where only one may be set.
warnUnlessAny	[{ "keys": [...], "message": "..." }]: preflight warns when none of the keys is set.
signIn	{ "mode": "chrome" }, { "mode": "none" }, or { "mode": "form", "login": { "url", "fields": { "<selector>": "<ENV_VAR>" }, "form", "success" } }. A form sign-in only names environment variables; values never appear in a file.
css	CSS added before every shot. Hide notices, counts, banners, and first-visit dialogs here rather than clicking them away: a click can save a preference to the account.
release	{ "source": "..." }, the default for release-check --config. See Commands.
actions	Named steps and checks. Each is { "params", "steps", "expect", "expectAfter" }; {param} in any string is replaced. Steps: { "click": selector, "text"?, "frame"?, "skipIf"?, "optional"? }, { "scroll": selector }, { "wait": selector or ms }, { "key": "Escape" }, { "moveMouse": { x, y } }. Checks: a selector, or { "selector", "text"?, "frame"?, "absent"? }; a shot fails and saves nothing when a check fails.

Check selectors against the real app before you rely on them, and add a starter brief in briefs/<preset>.json. The preset tests (node tests/presets-tests.mjs) load every preset, refuse a step that names a destructive control, and dry-run every starter brief.