agentleFS
chanktbchanktb/claude-shopify-growth/skills/product-content-deep/SKILL.md

product-content-deep

Skill2 starsChanged 5 months agoMITView on GitHub

Generate a unique, benefits-first product-page (PDP) description for a Shopify product (7 tier-aware sections: Lead, Key Benefits, Specs Table, How to Use, Best For / Use Cases, Comparison, FAQ). Replaces manufacturer boilerplate with original copy that satisfies the product-analyze Category 1 (Product Copy Quality) and Category 5 (GEO/AI-Citation) rubrics. Outputs an HTML body for `descriptionHtml` + a flat FAQ array for `FAQPage` schema. Pairs with product-analyze (audit) and product-optimize (push + schema). Use when user says "write product description", "product copy", "PDP content", "rewrite product listing", "unique product description", "product-content-deep".

The file

Show its rules and text

We split this skill into the rules it gives an agent

15 rules. About 0 are common, 9 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈9

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If none of the six signals is clear, default to a balanced weighting across all seven sections at the mid-length tier (350w).

    line 82
  2. Distinctive:

    No AI-trigger phrases (see Em-Dash + AI-Trigger Compliance below)

    line 100
  3. Distinctive:

    No bullet duplicates the Lead sentence

    line 121
  4. Very specific:

    Before writing ANY copy, grep your product knowledge base at knowledge-base/*.md for this product line.

    line 214
  5. Very specific:

    Max Paragraph Length: every <p> MUST be short (2–4 sentences max, or ≤80 words).

    line 263
  6. Distinctive:

    No paragraph exceeds 4 sentences or 80 words

    line 270
  7. Distinctive:

    No AI trigger phrases: "discover", "elevate", "seamless", "curated", "delve into", "in conclusion", "navigate the landscape", "in today's …

    line 271
  8. Distinctive:

    No internal source labels: never write "per our knowledge base", "the KB", "internal data" in customer-facing copy.

    line 275
  9. Distinctive:

    Vocabulary diversity: don't reuse the same adjective ("great", "amazing") across sections.

    line 287
  10. Distinctive:

    No formulaic scaffolding: avoid "Not only... but also", "Whether you're...

    line 289
  11. Very specific:

    Does NOT push content to Shopify (that's product-optimize, which also runs the HARD gate sweep before publish)

    line 339
  12. Very specific:

    Does NOT generate or edit product images (use the blog-image skill or a dedicated image pipeline)

    line 341
  13. Very specific:

    Does NOT audit or score a PDP (that's product-analyze.

    line 343
  14. Distinctive:

    Does NOT modify price, inventory, or variant data

    line 345
  15. Very specific:

    Does NOT emit the Product/Offer/AggregateRating schema itself (that bundle is built in product-optimize Step 8, this skill only supplies …

    line 346

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Product Content Deep: Benefits-First PDP Description Generator

Generates the unique, benefits-first product description that replaces manufacturer boilerplate on a Shopify product page (PDP). Solves the #1 failure mode flagged by product-analyze Category 1 (Product Copy Quality): pasting the manufacturer's thin, duplicated copy instead of writing an original, buyer-focused description. This skill brings PDP copy depth on par with a well-scoped landing page while keeping the buy-box the primary conversion element (description supports the sale, it doesn't replace the buy box).

When this skill applies

Any product page missing a unique description, flagged by product-analyze Category 1 < 15/20 or a GATE 6 hit ("description duplicated across > 5 products"). Applies to a single product per invocation; batch it from product-audit-pipeline or product-optimize for multiple SKUs.

Tier requirement: none, this skill can run on any product. The tier (see below) changes section emphasis and length, not whether the skill applies.

Architecture

Content is generated as an HTML body for the product's own descriptionHtml field (NOT a metafield, unlike the collection module) plus a companion flat FAQ array pushed to a metafield that feeds FAQPage schema.

Product.descriptionHtml
  └─ Lead + Key Benefits + Specs Table + How to Use + Best For + Comparison

Product.metafields.custom.faq_jsonld  (flat [{q,a}], same discipline as collections)
  └─ rendered as an on-page Q&A block + FAQPage JSON-LD by the theme

This pairs with the schema bundle in ../product-analyze/references/product-schema.md — the specs table section below must use the same names/values as the additionalProperty block, and the FAQ array must stay flat (never a schema.org wrapper) so the theme can build both the visible accordion and the JSON-LD from one source.

Category Taxonomy (tier-aware emphasis)

Detect (or accept via --tier) which of these six generic-ecommerce categories the product belongs to. Each shifts section depth and the unique-copy bar:

TierDetection signalEmphasis
variant-rich≥5 variants (size/color/style), e.g. apparel, footwearSpecs table covers the variant matrix; Comparison section leans on "which variant/size to pick"
technical-hardwareSpec-driven durable, e.g. electronics, tools, appliancesDeeper Specs Table + Comparison (vs. the next model up or a competitor spec); How to Use is mandatory (setup steps)
bundles-setsMulti-item kit or set, e.g. a starter set, gift bundleKey Benefits emphasizes "what's in the box" + value vs. buying separately; Comparison = bundle vs. à la carte
consumables-bulkLow-differentiation, repeat-purchase, e.g. refills, pantry goodsRelaxed unique-copy bar (short Lead + Benefits is fine); emphasize value/reorder cadence and pack-size options over deep specs
seasonal-dropsTime-boxed or limited-run item, e.g. holiday decor, limited colorwayLead emphasizes timing/availability; Best For leans on gifting/occasion use cases
specialty-premiumHigher price point, niche audience, e.g. premium home goods, boutique cosmeticsDeepest Lead + Comparison (justify the price tier); Best For narrows to the specific buyer profile, not a general audience

If none of the six signals is clear, default to a balanced weighting across all seven sections at the mid-length tier (350w).

7 Sections (default template)

Each section is a heading + body inside descriptionHtml, except the FAQ, which is a separate flat array. Total target: 200–500 words in the description body depending on --length and tier (technical-hardware and specialty-premium default to 500w; consumables-bulk defaults to 200w).

Section 1: Lead (1–2 sentences, ~20–40w)

The outcome or core use up front, not the SKU or brand recitation. Answer "what does this do for me" before "what is this."

Required elements:

  • Opens with the benefit or use case, not the product name repeated from the H1
  • No AI-trigger phrases (see Em-Dash + AI-Trigger Compliance below)

Anti-pattern: "The Acme Wireless Charger is a wireless charger made by Acme that charges your phone wirelessly." Rewrite: "Drop your phone on this pad and it starts charging in under two seconds, no cable, no fumbling in the dark."

Section 2: Key Benefits (3–5 bullets, ~60–100w)

Feature → outcome bullets, not a spec dump. Each bullet leads with the benefit, then names the feature that delivers it.

Format:

<ul class="pdp-content__benefits">
  <li><strong>{Benefit}</strong>: {feature that delivers it, 1 sentence}</li>
</ul>

Required elements:

  • 3–5 bullets, each a single sentence
  • No bullet duplicates the Lead sentence
  • At least one bullet ties to a buying-decision concern (durability, compatibility, ease of use, value)

Section 3: Specs Table (structured, mirrors additionalProperty)

A semantic <table> with the same names and values you'll push to the additionalProperty schema block (per ../product-analyze/references/product-schema.md §3). Keep them in sync; a mismatch is a citability + trust gap, not just a schema gate.

<table class="pdp-content__specs">
  <thead>
    <tr><th scope="col">Spec</th><th scope="col">Value</th></tr>
  </thead>
  <tbody>
    <tr><th scope="row">Material</th><td>{value}</td></tr>
    <tr><th scope="row">Dimensions</th><td>{L x W x H}</td></tr>
    <tr><th scope="row">Weight</th><td>{value}</td></tr>
    <tr><th scope="row">Compatibility</th><td>{value}</td></tr>
  </tbody>
</table>

Rows vary by category (apparel: material/fit/care; electronics: power/ports/compatibility; home goods: dimensions/material/capacity; cosmetics: volume/ingredients-summary/skin-type). Never invent a row value; pull from the KB or Shopify metafields (see Factcheck Discipline below).

Section 4: How to Use / Setup (where relevant, ~50–100w)

Short ordered steps for first use, setup, or application. Mandatory for technical-hardware; optional (skip if truly not applicable) for consumables-bulk and bundles-sets where "use" is self-evident.

<ol class="pdp-content__howto">
  <li><strong>{Step name}</strong>: {1 sentence}</li>
</ol>

Section 5: Who It's Best For / Use Cases (~40–80w)

States who/what it's best for, mapping to long-tail "best X for Y" AI queries. 2–4 short use-case statements, not a generic "everyone will love this."

Format: short paragraph or a 2–4 item <ul>, e.g.:

<ul class="pdp-content__best-for">
  <li>{Use case 1}: {why it fits}</li>
  <li>{Use case 2}: {why it fits}</li>
</ul>

Section 6: Comparison (~40–100w, prose or table)

How this product compares to alternatives or the next variant up. Tier shapes the comparison target:

  • variant-rich: this size/color vs. the others in the same line
  • technical-hardware: this model vs. the next tier up, or vs. a common competitor spec point (no disparaging claims, factual deltas only)
  • bundles-sets: bundle price/value vs. buying items à la carte
  • consumables-bulk: this pack size vs. other pack sizes (value per unit)
  • seasonal-drops: this drop vs. the standard/year-round version
  • specialty-premium: this tier vs. the brand's entry-level option, justifying the price gap with concrete differences

Can skip with --no-comparison when there is truly no valid comparison target (single-SKU brand-new product with no line siblings).

Section 7: FAQ (3–5 Q&A, flat array, NOT part of descriptionHtml)

Real buyer questions (compatibility, sizing, use, care, returns), written as a flat array, never a schema.org wrapper:

[
  { "q": "{buyer question}", "a": "{direct answer, 1-3 sentences}" },
  { "q": "{buyer question}", "a": "{direct answer, 1-3 sentences}" }
]

--faq-count controls 3–5 (default 4). The theme renders the visible accordion and the FAQPage JSON-LD from this same flat source (see ../product-analyze/references/product-schema.md §5).

Generation Process

Step 0.5: MANDATORY knowledge-base factcheck (HARD GATE 0)

Before writing ANY copy, grep your product knowledge base at knowledge-base/*.md for this product line. Extract real specs, materials, compatibility, use cases, and differentiators. Never invent a spec. If the KB doesn't cover it, pull from Shopify product metafields, then the manufacturer's spec page (WebFetch), and log a gap rather than guessing. Same TIER_INTERNAL → TIER_SHOPIFY → TIER_WEBFETCH discipline as collection-content-deep and product-optimize.

Step 1: Pull product data

Pull via the Shopify connector:

  • Title, handle, existing descriptionHtml (source of any real facts worth keeping; do not blindly discard), productType, vendor, tags
  • Variants: price, compareAtPrice, sku, barcode, options (informs the Specs Table + variant-rich Comparison)
  • Images + existing alt text
  • Existing metafields: custom.faq_jsonld if present (don't duplicate themes already covered), any specs/additionalProperty metafields already populated

Step 2: Detect tier

Auto-detect per the Category Taxonomy table above, or accept --tier override. This decides section emphasis, default length, and how strict the unique-copy bar is (relaxed for consumables-bulk, strictest for specialty-premium and technical-hardware).

Step 3: Identify comparison target

For variant-rich: pull sibling variants/options. For technical-hardware and specialty-premium: identify the next tier up or the most common alternative named in the KB. For bundles-sets: pull the à la carte prices of the bundle's components. Skip per --no-comparison if none applies.

Step 4: Generate per-section content

For each of the 7 sections, generate text using:

  • KB facts (Step 0.5) — never invented specs
  • Store-specific facts (pricing, shipping, returns) only where relevant to Key Benefits or Comparison
  • Tier-specific emphasis (Category Taxonomy table)

Step 5: Em-dash + AI-trigger compliance & Readability Constraints

Rule: keep em-dash usage very low, ≤0.3 em-dash per 100 words across ALL copy (description + FAQ). Use colons / periods instead. Run scripts/reduce_emdash.py before finalizing.

Readability & Paragraph Rules (CRITICAL for mobile UX):

  • Max Paragraph Length: every <p> MUST be short (2–4 sentences max, or ≤80 words).
  • Scannable Layout: bullets for benefits/best-for, a semantic table for specs, bold lead-ins for key points. No wall-of-text paragraphs.

Self-check before push:

  • Count em-dashes in generated body + FAQ (ratio ≤0.3/100w)
  • No paragraph exceeds 4 sentences or 80 words
  • No AI trigger phrases: "discover", "elevate", "seamless", "curated", "delve into", "in conclusion", "navigate the landscape", "in today's fast-paced", "it's important to note", "let's explore", "experience the difference"
  • No internal source labels: never write "per our knowledge base", "the KB", "internal data" in customer-facing copy. State facts directly as the store's own product knowledge.

Anti-AI-detection (human voice) — the trigger blocklist alone is not enough; the stronger tells are rhythm and vagueness:

  • Burstiness: vary sentence length. Mix short, punchy lines with longer ones; uniform same-length sentences read as machine-generated.
  • Concrete over vague: replace generic claims ("high quality", "perfect for everyone", "premium materials") with a specific KB fact (the actual material, the measured capacity, the real use case). Specificity is the #1 human signal on a PDP and also the thing a buyer actually needs.
  • Vocabulary diversity: don't reuse the same adjective ("great", "amazing") across sections; don't open every bullet the same way.
  • No formulaic scaffolding: avoid "Not only... but also", "Whether you're... or...", and "Looking for X? Look no further" openers.

Step 6: Unique-copy check (HARD GATE, ties to product-analyze Cat 1.3)

Compare the generated description against the manufacturer's copy (if known) and against your own other product descriptions. Target: <70% similarity to manufacturer boilerplate, and not duplicated across >5 of your own products. consumables-bulk gets a relaxed bar: near-identical low-differentiation items (e.g. the same formula in different pack sizes) can share most of the Lead + Benefits, but the Specs Table and pack-size Comparison must still be unique per listing.

Step 7: HTML structure

Wrap each descriptionHtml section in:

<section class="pdp-content__section">
  <h2 class="pdp-content__heading">{Section Title}</h2>
  <div class="pdp-content__body">
    <!-- section content -->
  </div>
</section>

FAQ is written separately as faq.json (flat array), not inside descriptionHtml.

Step 8: Verified internal links (where the section calls for one)

Any link into a parent collection, a sibling product, or a blog post named in Best For / Comparison must be confirmed against a live URL inventory pull first. Never invent a /products/, /collections/, or /blogs/articles/ path.

Composition with other skills

  • Before: product-analyze (identifies which PDPs need a description rewrite and which tier applies)
  • After: product-optimize (pushes descriptionHtml + FAQ metafield + schema bundle live, runs the pre-publish gate sweep, re-scores)
  • Schema ref: ../product-analyze/references/product-schema.md (specs table must mirror additionalProperty; FAQ must mirror FAQPage)
  • Reads from: knowledge-base/*.md (product facts), existing descriptionHtml (salvageable real facts)
  • Writes to: description.html + faq.json per product (consumed by product-optimize Step 6/9, not pushed directly by this skill)

What this skill does NOT do

  • Does NOT push content to Shopify (that's product-optimize, which also runs the HARD gate sweep before publish)
  • Does NOT generate or edit product images (use the blog-image skill or a dedicated image pipeline)
  • Does NOT audit or score a PDP (that's product-analyze; run it first to confirm this skill is needed, and after to confirm the lift)
  • Does NOT modify price, inventory, or variant data
  • Does NOT emit the Product/Offer/AggregateRating schema itself (that bundle is built in product-optimize Step 8, this skill only supplies the specs table and FAQ data it consumes)

Output convention

Per-product artifact:

output/[site]/products/[handle]/
├── description.html      ← HTML body for descriptionHtml (this skill's output)
└── faq.json              ← flat [{q,a}] array for the FAQ metafield (this skill's output)

This mirrors the output/[site]/products/[handle]/ convention already used by product-analyze and product-optimize, keeping the ledger format consistent across the kit. product-optimize reads both files directly in its Step 6/9.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.