Skills for building AEM Edge Delivery Services sites — site creation, block development, content modeling, code review, testing, and page import.
91%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
Low
Low-risk findings worth noting
Status: proposed — optional. This spec is not required for the skill to work: infer-and-confirm (SKILL.md Phase 2) is the primary path and needs no annotations. This contract is a good-to-have for teams that want to pre-declare section mappings and skip the confirmation round-trip. Everything below is a recommended starting spec to formalize with adopters if/when annotations are adopted.
Annotations are the high-confidence path. When the design tells the skill, per section, which block it becomes and which content goes where — and, when a section needs a block the project doesn't have, that it should be built — the mapping is authoritative and no inference is needed.
When annotations are absent, the skill does not stop; it falls back to inference + confirmation (SKILL.md Phase 2): it matches each section against the project's existing block palette on structure and visual fit, then presents the resulting plan for the user to confirm, asking on any low-confidence section. Annotating removes that guesswork and the confirmation round-trip — prefer it for reliability and unattended automation.
For each logical section of the page frame, a section annotation is recommended. A section with no annotation is not dropped — it goes to inference + confirmation (SKILL.md Phase 2.1–2.2): the skill proposes a mapping and confirms it with the user, asking when the section is ambiguous. Annotating simply makes that section high-confidence and skips the confirmation round-trip.
A "logical section" is not always a direct child of the frame. SKILL.md's segmentation heuristic merges sibling nodes (e.g. a heading node, a divider, and a row of cards laid out as separate siblings) into one visual band that becomes a single EDS section. The annotation attaches to that logical section, not necessarily to a single direct child:
Optional page-level annotations:
metadata block; see da-content SKILL.md).section-metadata block; see da-content SKILL.md).get_design_context / get_metadata).If both are present, the Dev Mode annotation wins.
key: value block)type: existing-block # existing-block | new-block | default-content
block: cards
variant: highlight
fields:
- title: "Layer: Card Title"
- description: "Layer: Card Body"
- image: "Layer: Card Image"
- cta: "Layer: Learn more"| Key | Required | Meaning |
|---|---|---|
type | yes | The section outcome — one of existing-block (author into a block the project already has → Phase 3A), new-block (build a new isolated block, content+code → Phase 3B), or default-content (prose/media directly in the section, no block → Phase 3C). |
block | block types only | Target block name. Required when type is existing-block or new-block; must be omitted when type: default-content. |
variant | no | Block variant(s), rendered as extra class tokens (cards highlight). Block types only. Must be a variant the block defines (existing-block) or one the new block will define (new-block). |
fields | no | Explicit map of block field → Figma layer. Block types only (ignored for default-content). When omitted, field mapping is inferred from the block's content model + the section's visual order. Provide fields when inference is ambiguous. |
new | no (deprecated) | Legacy discriminator, superseded by type: new: true ≡ type: new-block; new: false or absent (with a block) ≡ type: existing-block. Accepted for back-compat; type wins if both are present. new cannot express default content — that is exactly why type was added. |
#hero
block: cards (highlight)#name is treated as block: name with type: existing-block.block: name (variant) sets block + variant inline.!new to request creation — block: pricing-table !new ≡
type: new-block.#default (or type: default-content) marks the section as
default content — no block, prose/media authored directly (Phase 3C).fields map in shorthand — mapping is inferred.When fields is omitted, map content to the block's content model using:
Ambiguity (more content nodes than cells, or types that don't line up) is an
error, not a guess: report the section as needing an explicit fields map.
type: existing-block — block exists and its look fits the section →
content-only. Author content into it (Phase 3A). If the block matches
structurally but its styling diverges (only its own CSS could produce the
look), it's a new block instead — Phase 3A reuse gate.type: new-block (or legacy new: true, or a block that names something
absent, user-confirmed) → content+code. Create it as a new isolated block
(Phase 3B); never skin existing blocks (project-level token theming aside —
see SKILL.md Guardrails).type: default-content → no block. Author prose/media directly in the
section <div> (Phase 3C); a block/variant/fields alongside it is a
malformed annotation — report and ask, don't guess which the user meant.type: existing-block names something absent → confirm, don't assume.
Ask the user (build a new block under this name, or did they mean an existing
one?) before proceeding — a malformed annotation is treated as low-confidence,
not silently built.Figma frame Spring Campaign with five sections:
| Section | Annotation | Resolves to |
|---|---|---|
| Hero banner | block: hero (type: existing-block) | content-only → <div class="hero"> |
| 3 feature cards | block: cards (highlight) | content-only → <div class="cards highlight">, one row per card |
| Intro paragraph + heading | #default (type: default-content) | default content → <h2>/<p> directly in the section <div>, no block (Phase 3C) |
| Interactive pricing table | block: pricing-table !new | content+code → build a new isolated pricing-table block, then author content |
| Newsletter strip | (none) | inferred → matched against the palette, proposed in the plan, confirmed with the user |
Result: hero + cards author immediately; pricing-table is built as a new
block and its code pushed before preview; the newsletter strip is inferred
(Phase 2) — the skill proposes a mapping (reuse an existing block, default
content, or a new block) and confirms it with the user before building, rather
than being skipped.
variant is expressed (Figma component variants vs. a free-text
annotation key). (The existing-block / new-block / default-content
distinction is now carried by the type key — §3a; legacy new is a
deprecated alias.)type reliably for adopters..tessl-plugin
skills
aem-block-library
aem-cli
aem-layout-view
block-collection-and-party
brand-setup
references
building-blocks
code-review
content-driven-development
references
content-modeling
create-site
da-auth
figma-to-content
find-test-content
generate-import-html
identify-page-structure
migrate-header
page-decomposition
page-import
preview-import
scrape-webpage
slicc-handoff
snowflake
assets
substrate
substrate-milo
examples
knowledge
scripts
testing-blocks
ue-component-model