CtrlK
BlogDocsLog inGet started
Tessl Logo
PluginPublic

adobe/stardust

Redesign an existing website to make it better. Built on top of impeccable.

Quality

88%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Low

Low-risk findings worth noting

Overview
Quality
Evals
Security
Files

config.mdskills/spec/reference/

spec.config.json

stardust/spec/spec.config.json, committed with the output (--config names another file). Paths resolve against the project root; judgement and knowledge go under dir, the run's raw material under work. Keys not listed here are ignored.

keyrequiredmeaning
siteyesdisplay name, written to knowledge/site.json
originyeshttps://<host> after redirects (S1 stops and names the final origin otherwise), no trailing slash
scopePathyespath prefix of the scope, e.g. /en/ or /<country>/<lang>/
dirnooutput folder (judgement/, knowledge/), default stardust/spec
worknoraw material of the run, default stardust/.work/spec (never committed)
sitemapsnositemap URLs; default: robots.txt Sitemap: lines, else /sitemap.xml; none in scope → a link crawl from scopePath
maxPagesnocap on the URLs every stage covers; above it S1 keeps an even sample per section and the spec says so
workersnoparallel requests for fetch, default 4 (be polite: the source is production); --headed runs one page at a time, --archive two
templateyeshow a page names its template: { "bodyAttr": "data-template" }, { "bodyClass": "^editable-(.+)$" } (one capture group), or { "meta": "<name>" }; "pathSegments": N appends the first N path segments under scopePath (alone, it is the whole template name)
parseryes{ "profile": "aem-classic" | "aem-core" | "generic", "main": "<selector>", "chrome": { "header": "<selector>", "footer": "<selector>" }, "nameAttrs": [], "stripPrefix": "<regex>", "extraRoots": [] }; nameAttrs and stripPrefix only for generic, extraRoots only for aem-classic
rumno{ "keyEnv": "RUM_DOMAIN_KEY", "domain": "<host>", "days": 90 }; omit or null when no key exists
referenceBlocksno{ "name": "<library>", "count": <n>, "source": "<where the blocks live>" } — what reuse verdicts compare against

Choosing the template rule

Open two pages of different kinds and read <body>. Classic AEM pages often carry data-template; editable-template sites carry the template in a body class; other CMSs expose a meta tag or a body class prefix. When nothing names the template, use { "pathSegments": 1 } and let S8 variants do the grouping.

Choosing the parser profile and main selector

  • aem-classic: custom components with a c- class prefix inside div.main.content, column controls (colctrl); responsive-grid members are named by the component they wrap; extraRoots lists component classes without the prefix.
  • aem-core: Core Components / editable templates — components are responsive-grid members (aem-GridColumn); main is usually the page's main container (#mainContent, main, .root > .aem-Grid).
  • generic: any other site. Below main, single-child wrappers are skipped and the children are the components, named from a component attribute, a block-style class (wp-block-*, elementor-widget-*, paragraph--type--*), the block of a BEM or CSS-module class, else a shape label (section.h2.list, div.media) where classes are only utilities. Without main, the chain main, [role=main], #main, #content, #main-content, article, then the body without header/footer/nav. Tune with nameAttrs (attributes the site uses for component names) and stripPrefix (a hash prefix that makes the same component look different per page).
  • Check with spec-parse.mjs --fetch <40 rows> and spec-profile.mjs: the vocabulary must be component types, not style classes, and has_main true on every page. A new CMS family is a new profile in lib.mjs profileRules (a root rule or structural roots, optional columns and row widths): spec-parse and the spec-capture tagger both run it through componentLayout, so path ids cannot drift; the node contract in spec-parse's header does not change.

skills

README.md

tile.json