CtrlK
BlogDocsLog inGet started
Tessl Logo
PluginPublic

adobe/aem-edge-delivery-services

Skills for building AEM Edge Delivery Services sites — site creation, block development, content modeling, code review, testing, and page import.

Quality

91%

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

README.mdskills/snowflake/

snowflake

A skill for converting AI-generated static HTML pages (Stardust, Mobirise, Relume, Lovable, v0, Figma-derived hand-coded, etc.) into Adobe Edge Delivery Services pages using the overlay pattern.

The overlay pattern preserves the original DOM byte-for-byte. Only the text and image content becomes authorable in Document Authoring. Header and footer remain static repository fragments. The page CSS and any animation JavaScript ship per-template under the EDS code bus.

This skill encodes the 6-phase methodology distilled from 5 R&D iterations against varied generator outputs and bespoke pages.

What's included

skills/snowflake/
├── SKILL.md                       Entry point (agent reads first)
├── phases/
│   ├── 1-capture.md
│   ├── 2-analyze.md
│   ├── 3-generate.md
│   ├── 4-wire.md
│   ├── 5-roundtrip.md
│   └── 6-reflect.md
├── knowledge/
│   ├── methodology.md             Canonical phase rules
│   ├── architecture.md            Overlay engine + slot writer reference
│   ├── eds-da-mechanics.md        EDS pipeline reference (overlay-runtime lore; DA HTML and admin API in da-content skill)
│   └── learnings.md               Cross-project findings (5 runs)
├── scripts/
│   ├── transform-da-to-eds.mjs    DA divs-with-class → drafts HTML
│   └── dom-equality.mjs           Source vs rendered DOM comparison
├── examples/
│   └── README.md                  Pointers to worked examples
├── HOST-NOTES.md                  Per-host adapter notes (Slicc, Claude Code, generic shell)
└── README.md                      You are here

Prerequisites

The target EDS repo must have the overlay substrate in place: scripts/overlay-engine.js with applyTemplateOverlay, the writeSlot slot writers, and related template/slot logic. scripts/scripts.js is Adobe's standard boilerplate lifecycle file (loadEager/loadLazy/etc.) and receives only a small injected hook — an import of overlay-engine.js and an overlay guard at the top of loadEager that early-returns for overlay pages, skipping EDS decoration entirely. Also required: the blocks/header/header.js / blocks/footer/footer.js fetch decorators. See knowledge/architecture.md for the design.

The skill installs this for you. phases/0-prereq.md drives an idempotent installer (install-substrate.mjs) that brings a vanilla adobe/aem-boilerplate clone up to overlay-capable. Run it once per target repo:

node <SKILL_DIR>/scripts/install-substrate.mjs --dry-run   # inspect plan
node <SKILL_DIR>/scripts/install-substrate.mjs             # apply

Overwritten files are backed up to .snowflake/.backup/<timestamp>/. Subsequent runs of the skill detect the installed substrate via .snowflake/config.json and skip Phase 0 silently.

See HOST-NOTES.md for per-host details (Slicc, Claude Code, generic shell).

Limitations of v1

This is the minimum-viable version. It runs the 6 conversion phases sequentially, no parallel fan-out. Future-work items:

  • Subagent fan-out at Generate (~4-6 min wall-clock savings)
  • Generate-time validators (catches nested [data-slot] and non-absolute DA cell <img> URLs before round-trip)
  • Spec-mode Analyze (structured decisions.json schema validation)
  • Round-trip failure-triage subagent (Opus + extended thinking)

Installation

See HOST-NOTES.md for the install command and configuration per host (Slicc, Claude Code, generic shell).

Contributing

The methodology, architecture, and learnings in knowledge/ are deliberately bundled with the skill, not referenced from external sources. This keeps the skill portable and stable.

Users running snowflake against new sources accumulate project-specific learnings in their own repos (under .snowflake/projects/<NNN>-<slug>/learnings.md and optionally .snowflake/knowledge/learnings.md for the repo-level pile). When a finding is generic enough to benefit other snowflake users — a new pipeline behavior, a new slot writer pattern, a class of bug worth catching — please raise a PR to this skill repo with the finding promoted into knowledge/learnings.md (and any corresponding rule into knowledge/methodology.md).

Convention for learnings.md entries: dated ## YYYY-MM-DD — short title at the top of the file. Each entry: brief context, observable symptom (if a bug), the rule.

Status

v1 — minimum viable. Bundles the methodology and architecture distilled across 5 runs. Single-sequence (no parallelism). Designed to run on any host that provides bash + node + git + curl + playwright-cli.

Open this skill's roadmap and known issues in the upstream PR.

skills

README.md

tile.json