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

knowledge.mdskills/spec/reference/

The spec's output (knowledge/)

spec-knowledge.mjs writes stardust/spec/knowledge/ from the run's raw material and judgement/; clients read it, never crawl or judge. Rewritten whole on each run, rows in a stable order, names neutral. A patch only adds fields or enum values, which clients ignore when unknown; removing or renaming one is a minor release.

Files

url_id is the id of a urls.jsonl row.

FileRows and main fields
site.jsonsite, origin, scope_path, reference_blocks {name, count}, rum {available, window, bundles}, sitemap_urls, inventory_total, inventory_source, sample_note (only when capped), fetch_sources {live, headed, archive}, evidence_note (only with archive captures), built_at, built_by, provenance, variant_cut, capture_format, i18n_notes[]
urls.jsonlid, url, path, section, depth, in_sitemap, status, final_url, final_status, outcome, template, variant_code, title, eds_path, needs_migration_redirect, pageviews_90d, rum_bundles, traffic_band, block_count, capture_key, main_chars, flag, blocked (why an error row was not read)
page-blocks.jsonlper page: url_id, url, blocks[] {pos, block, variant, kind (block, dynamic, widget, default, metadata, global), source[], path, nested_in, section, crop}; chrome globals first and last
signals.jsonlper live sitemap page: url_id, url, signals[] (hreflang, script:<host>, iframe:<host>, jsonld:<type>, and implementation.json#signals names)
blocks.jsonname, kind (`block
variants.jsoncode, template_id, rank, label, core[], optional[], url_count, pageviews_90d, distinct_sets, rep_url, rep_capture_key
templates.jsonid, label, url_count, pageviews_90d, variant_count, top_variants_share, rep_url
source-components.jsonname, url_count, instance_count, maps_to {block: instances}
redirects.jsonlsrc, target, status, hops, kind (legacy: redirects today; migration: every changed delivered path, .html included), needs_entry (migration: 1 when the path changes beyond the .html, index-name and trailing-slash rule; null on legacy), target_status, external, in_sitemap, inbound_pages, rum_views, note
broken.jsonlurl, status, source (sitemap / link / rum), kind, in_scope, inbound_pages, inbound_main, rum_views, rum_bundles, referrers, note
bad-links.jsonlper dead or redirected target: to_url, main[] and chrome[] (url ids of the pages linking to it)
redirect-landings.jsonpath, views, bundles
features.jsonid, class, class_name, name, evidence, disposition, reproducibility, status, pattern, eds, decisions[], reach, sitewide, reach_pages, reach_templates, pageviews_90d, pages[] (empty when site-wide)
martech.jsonconsent_summary, loading_order[] {stage, what}, vendors[], launch_rules[], datalayer[], evidence {launch, onetrust, consentMode} (spec-martech's files, unchanged)
metadata.jsonname, source, used_by[], coverage_pages, distinct_values, top_values
query-indexes.json, helix-query.yamlname, include_paths, exclude_paths, filter, properties, consumers, source, yaml; the yaml file joins them (dynamics skeleton)
locales.jsontree, country, language, urls, rum_views_90d, sitemap, …
site-config.jsonkey, now, eds, decision
archetypes.jsoncut, types[] {type, template, variant_codes[], rep_url, urls[], url_count}: one page type per variant covering cut of its template
header.jsonurl, at, summary[]: the header's size per width (controls, states, links, depth), from S9's header exploration
search-probes.jsonterm, expect_count, expect_titles[], expect_includes, feature, path, param
open-questions.jsonid, area, owner (stakeholder / implementer), blocking (1: confirm the default before launch; a different answer means rework), question, context, options[], default_assumption, default_option (index of the option the default is, or null), impact_rule, impact, link, features[], answer {answer, option, by, at}, effective
findings.json{ title, text, numbers[], chapter, rank }, plain text, every number computed; chapter of the spec's story (start, pages, build, dynamic, risks, decide), rank 1 leads it

Rules

Feature reach, question impact and {{…}} in findings use one rule format, evaluated by rules.mjs. A rule that cannot be computed stops S10 with its text; there is no fallback to a typed number. Extend the format in rules.mjs when a real run needs a figure it cannot express.

  • Filter: field=v[,v…] or field!=… on a row; * is a case-insensitive glob; null matches a missing value; quote values with spaces or commas (template="page · news").
  • URL terms (reach, urls …): live (outcome page), sitemap, nested, block:<b>[|<variant>][,…], signal:<glob>[,…], verdict:<v> (a block on the page has that verdict), or any filter on a urls.jsonl row; ! negates; terms are ANDed. Example reach: live sitemap signal:script:*player*,iframe:*player*.
  • Numbers: urls <terms…> · count <set> [filters…] · sum <set>.<field> [filters…] · <set>[<key>].<field> · <set>[max:<field>].<field> · value:<n>. Sets: urls, page-blocks, blocks, block-variants, variants, templates, features, redirects, broken, bad-links, vendors, launch-rules, datalayer, metadata, query-indexes, locales, site-config, search-probes, open-questions, source-components.
  • Examples: urls sitemap flag=empty · count redirects kind=legacy external=1 · blocks[cards].url_count.

skills

README.md

tile.json