Build a section-by-section inventory of your website content with quality scores across five dimensions: Clarity, Consistency, Depth, CTA Strength, and Trust Signals
The AI confirms the scoring framework, then you submit pages in batches of up to 15, organized by site section. For each batch, the AI validates input, builds a cumulative site inventory with inferred page types and purposes, and produces a section scorecard with justified scores. Starting from the second section, the AI identifies overlaps, contradictions, and emerging patterns across sections. After all sections are complete, you receive a full comparative analysis with section rankings, dimension-level insights, and a prioritized list of actionable improvement opportunities.
Website Quality Scorecard
Tags: website audit, content quality, scoring framework, site inventory, section scorecard, cross-section analysis, comparative analysis, improvement opportunities, interactive multi-turn, batch processing
TL;DR
What It Does
Builds a section-by-section inventory of your website content with quality scores across five dimensions: Clarity, Consistency, Depth, CTA Strength, and Trust Signals. You provide page data in batches organized by site section; the AI validates input, scores each section on a 0–5 scale per dimension, identifies cross-section patterns, and delivers a final comparative analysis with your top improvement opportunities ranked by impact. All scores are the AI’s subjective assessment of the content you provide — they create a consistent framework for comparing sections across your site, not absolute quality metrics.
How It Works
You provide your website name, business goal, target audience, and page data (URLs, titles, sections, and ideally content excerpts). The AI confirms the scoring framework, then you submit pages in batches of up to 15, organized by site section. For each batch, the AI validates input, builds a cumulative site inventory with inferred page types and purposes, and produces a section scorecard with justified scores. Starting from the second section, the AI identifies overlaps, contradictions, and emerging patterns across sections. After all sections are complete, you receive a full comparative analysis with section rankings, dimension-level insights, and a prioritized list of actionable improvement opportunities.
How To Start
STEP 1Set Your Scope
Tell the AI your website name, primary business goal, and target audience. The AI confirms the five-dimension scoring framework (Clarity, Consistency, Depth, CTA Strength, Trust Signals) and explains how scores work — they are structured subjective assessments, not measurements. You can add custom scoring dimensions at this stage if needed. The AI waits for your confirmation before proceeding.
Available parameters
-
website_name
· string · required
Name of the website to audit. -
business_goal
· string · required
Primary site objective. Be specific: “Convert visitors into demo requests for our project management tool” not “generate leads.” -
target_audience
· string · required
Who the site serves. Include role, company size, and what they are looking for. This calibrates whether content speaks to the right audience. -
page_data
· list · required
Page metadata: URL, title, section, and notes for each page. Organize by section. Include content excerpts (copy-paste key paragraphs) for more accurate scoring — without content, scores are marked [metadata-only] with lower confidence. -
custom_dimensions
· list · optional
Extra scoring dimensions beyond the five core. Only add dimensions you will actually act on — each one increases assessment time per page.
Example invocations
Solo founder auditing a small marketing site
#H->AI::Directive: (Execute Website Quality Scorecard with website_name: Acme Solutions, business_goal: convert visitors into free trial signups for our invoicing tool, target_audience: freelancers and small agency owners looking for simpler invoicing, page_data: [acme.com Home Product, acme.com/features Features Overview Product, acme.com/pricing Pricing Product, acme.com/about About Us Company, acme.com/blog Blog Content])
Marketing team assessing a multi-section SaaS site with custom dimension
#H->AI::Directive: (Execute Website Quality Scorecard with website_name: CloudStack Platform, business_goal: drive enterprise demo requests through content-led qualification, target_audience: mid-market IT directors evaluating infrastructure solutions, page_data: [cloudstack.com Home Product, cloudstack.com/platform Platform Overview Product, cloudstack.com/security Security Product, cloudstack.com/case-studies Case Studies Trust, cloudstack.com/pricing Pricing Product, cloudstack.com/docs Documentation Support], custom_dimensions: [SEO Optimization])
STEP 2Submit Page Batches
Provide pages in batches of up to 15, organized by site section. For each batch, the AI validates input (checking for duplicates, missing titles, and whether content excerpts are included), builds a cumulative site inventory with inferred content types and page purposes, and then scores the section across all five dimensions with a 1–2 sentence justification for each score. Pages without content excerpts are scored with a [metadata-only] qualifier and lower confidence. After each section scorecard, the AI waits for your review — you can challenge scores, provide additional context, or confirm and move on. Starting from the second section, the AI also reports cross-section signals: overlaps, contradictions, and emerging patterns, all backed by specific content citations.
STEP 3Review Your Scorecard
After all sections are complete, the AI delivers a final comparative analysis: the full cumulative site inventory, section rankings with strongest and weakest dimensions, a dimension-level analysis showing which dimensions are consistently strong or weak across your site, and a prioritized top 10 list of improvement opportunities ranked by estimated impact and effort. Every opportunity is actionable — not just “CTA is weak” but “add a specific benefit statement above the primary CTA button on the pricing page.” The analysis closes with site-wide observations and a methodology note reminding you to validate scores against actual user behavior data and analytics where available.
Usage Examples
Two abbreviated dialogues showing how the batch-based scoring and inventory building progresses. You provide page data and content excerpts; the AI validates, scores, and builds toward a comparative analysis.
Example 1 · Solo founder doing a first content audit of a small marketing site
User:#H->AI::Directive: (Execute Website Quality Scorecard with website_name: Acme Solutions, business_goal: convert visitors into free trial signups, target_audience: freelancers looking for simpler invoicing, page_data: [acme.com Home Product, acme.com/features Features Product, acme.com/pricing Pricing Product])
AI:I’ll score your Product section across five dimensions. Heads up: I don’t see content excerpts in this batch, so scores will be marked [metadata-only] with lower confidence. For better results next time, paste key paragraphs from each page. Here’s the scoring framework — does this match your assessment goals?
User:Looks good. Here’s the content from my home page: “Stop chasing payments. Acme sends invoices, tracks who’s paid, and reminds clients automatically. Free for your first 10 clients.”
AI:Much better — that home page content is strong. Clarity: 4/5 — the H1 directly states the benefit in language your freelancer audience uses. CTA Strength: 4/5 — “Free for your first 10 clients” reduces friction with a concrete threshold. Consistency: 3/5 [metadata-only for other pages] — can’t fully assess without features and pricing content. Product Section Total: 16/25 (64%). Want to challenge any scores before I move on?
[Session continues with the Company section. Cross-section signals reveal that the About page uses formal corporate language while Product pages use casual freelancer-friendly tone — a consistency gap. Final analysis ranks Product section strongest, Company section weakest, with “align About page voice to match Product section tone” as the top improvement opportunity.]
Example 2 · Marketing team with prior RCP-068 navigation audit results
User:We ran the Navigation Auditor (RCP-068) last week and found 3 dead-end pages and a confusing Resources dropdown. Here are our Product pages with full content excerpts. #H->AI::Directive: (Execute Website Quality Scorecard with website_name: CloudStack Platform, business_goal: drive enterprise demo requests, target_audience: mid-market IT directors, page_data: [6 Product pages with content])
AI:Good — I’ll factor your RCP-068 findings into Clarity and CTA Strength scoring. Those dead-end pages will likely show up as CTA Strength gaps. Validating your batch: 6 pages, all Product section, full content provided. No duplicates. Proceeding with scoring…
User:The Security page score seems low. We recently added SOC 2 certification details.
AI:Thanks for the context. Revising Trust Signals from 2 to 4 based on SOC 2 certification — that’s a strong trust signal for your IT director audience. Updated note: “Revised from 2 to 4 based on additional context: SOC 2 certification details recently added.” Section total moves from 17/25 (68%) to 19/25 (76%).
[After 4 sections, the comparative analysis reveals Trust section scores highest (case studies and certifications) while Support section scores lowest (documentation lacks depth for the technical audience). Top opportunity: “Add implementation guides with architecture diagrams to the documentation section to match the technical depth IT directors expect.” The AI notes these results can feed into a deeper RCP-072 Evidence Audit of the Support section.]
How AI Reads This Recipe
When this recipe is triggered, the AI acts as a structured content auditor building a cumulative site inventory with comparative quality assessments. The AI should:
- CONFIRM the scoring framework before any assessment begins. Present the five dimensions with their definitions and ask the user if they match the assessment goals. Do not score until the user confirms.
- VALIDATE each batch: check for duplicates, missing titles, and whether content excerpts are included. If no content is provided, flag that scoring will be [metadata-only] with lower confidence — but do not refuse to process the batch.
- SCORE honestly using the full 0–5 range. Every score must include a 1–2 sentence justification citing specific content or the absence of content. Avoid clustering everything at 3–4 — genuinely weak pages should score 1–2. If CTA Strength is not applicable for a page type (About, legal), score N/A and reduce the section maximum accordingly.
- WAIT for user review after each section scorecard. Users may challenge scores, provide additional context, or correct inferences. Adjust scores with revision notes when new context warrants it. Never argue about subjective scores.
- CITE specific content in cross-section signals. Do not make vague claims like “there seems to be inconsistency.” Point to specific pages and content elements on both sides of an overlap, contradiction, or pattern.
The AI should NOT invent page content, user behavior data, or performance metrics. It should NOT present scores as objective measurements — they are structured opinions. It should NOT assign scores without justification. It should NOT exceed 15 pages per batch. If the user shares prior RCP-068 Navigation Auditor findings, use them to calibrate Clarity and CTA Strength scoring but do not require prior navigation audit results.
When to Use This Recipe
Use this recipe when you:
- Need a structured assessment of your website content quality across multiple sections and want a consistent framework for comparing them.
- Want to identify which site sections are strongest and weakest to prioritize improvement work — especially useful before a site refresh or content overhaul.
- Are building a content inventory and want each page cataloged with its inferred type, purpose, issues, and improvement opportunities.
- Need to detect cross-section inconsistencies: messaging contradictions, tone mismatches, overlapping content, or site-wide patterns like weak CTAs everywhere.
Do not use this recipe when:
You need objective metrics like page load speed, conversion rates, or SEO rankings — this provides structured subjective assessment, not data-driven measurement. You want a structural navigation audit — use RCP-068 (Website Navigation Auditor) first for site architecture analysis. You need deep evidence-based strategic analysis of specific sections — use RCP-072 (Website Evidence Auditor) after this scorecard identifies which sections deserve deeper examination.
Recipe FAQ
Q.Are the scores objective measurements?
No. All 0–5 scores are the AI’s subjective assessment based on the content you provide. Their value is in creating a consistent framework for comparing sections across your site — not as absolute quality metrics. Think of them as structured opinions that help you see relative strengths and weaknesses. Validate against actual user behavior data and conversion analytics where available.
Q.How many pages can I audit at once?
Up to 15 pages per batch for quality. If your site has more than 15 pages, split them into batches organized by section (Product pages in one batch, Support pages in another). The inventory is cumulative — each batch adds to the running total. For very large sites (50+ pages), focus on the highest-traffic or most important sections first. You do not need to audit every page.
Q.What if I disagree with a score?
Challenge it. The AI waits for your review after each section scorecard specifically so you can push back. Share the additional context that changes your assessment and the AI will adjust the score with a revision note. Disagreement is expected and productive — you know your content and audience better than the AI does.
Q.Do I need to provide actual page content?
Not required, but strongly recommended. Without content excerpts, the AI can only assess based on URLs, titles, and section names — all scores will be marked [metadata-only] with lower confidence. Copy-pasting key paragraphs from each page (headlines, main body text, CTAs) gives the AI enough to produce meaningful scores. Even partial content for some pages is better than none.
Q.Should I run the Navigation Auditor (RCP-068) first?
Recommended but not required. RCP-068 provides a structural overview of your site navigation and identifies dead ends, confusing paths, and orphaned pages. Those findings help calibrate this recipe’s Clarity and CTA Strength scoring. If you have RCP-068 results, share them at the start of this session and the AI will incorporate them. If not, this recipe works independently.
Q.How does this connect to the Evidence Auditor (RCP-072)?
This recipe identifies WHERE to focus; RCP-072 goes deep on WHY. Use this scorecard’s section rankings and dimension analysis to decide which sections deserve the deeper evidence-based audit treatment. If a section scores consistently weak on Trust Signals and CTA Strength, RCP-072 can investigate why and produce strategic recommendations with supporting evidence. The three recipes in the Website Analysis series (068 → 071 → 072) move from structural overview to comparative scoring to deep strategic analysis.
Version History
Changes to this recipe over time. Most recent first.
v2.00a-R
2026-02-18
QA revision addressing 12 audit gaps (2 critical, 2 high). Added 11 behavioral rules covering score subjectivity, content requirements, honest justification, false-precision prevention, batch size enforcement, cumulative inventory maintenance, evidence requirements for cross-section signals, turn-taking, actionability, series awareness, and no-invention constraints. Added 4 WAIT gates for genuine multi-turn flow. Reframed scoring as subjective assessment with methodology notes. Added parameter coaching with examples. Reduced batch size from 20 to 15. Added [metadata-only] qualifier system. Updated series references to current IDs (068, 071, 072).
v2.00a
2025-12-31
Initial creation by Auguste. Interactive website content audit with section-by-section inventory and five-dimension scoring framework. Website Analysis series (2 of 3).
Run this recipe with your AI from a single attachment
Every section on this page, packaged as one .txt file your AI can consume directly. Download and attach.
THE ACTUAL RECIPE
RCP-000-000-071-WEBSITE-QUALITY-SCORECARD
Interactive website content audit that builds a section-by-section inventory with subjective quality assessments acrossfive dimensions (Clarity, Consistency, Depth, CTA Strength,Trust Signals). Produces comparative analysis and prioritizedimprovement opportunities.IMPORTANT: All scores are the AI's subjective assessment ofthe content you provide. They create a consistent frameworkfor comparing sections across your site — not absolute qualitymetrics. Validate against actual user behavior data andconversion analytics where available.
The CRAFT Recipe
# ===========================================================
# WEBSITE QUALITY SCORECARD
# Recipe ID: RCP-000-000-071
# Version: v2.00a-R (QA Revised)
# Series: Website Analysis (2 of 3)
# =========================================================== # ===========================================================
# BEHAVIORAL RULES
# ===========================================================
#
# RULE 1: SCORES ARE SUBJECTIVE ASSESSMENTS
# All 0-5 scores represent the AI’s subjective
# assessment based on available information, NOT
# objective measurements. Every scorecard must
# include a brief note: “Scores reflect AI
# assessment of provided content and may differ
# from user experience testing or analytics data.”
#
# RULE 2: NO SCORING WITHOUT CONTENT
# If the user provides only URLs and titles without
# content excerpts, the AI must flag that scoring
# accuracy is severely limited. Scores based on
# metadata only must be marked “[metadata-only]”
# and carry a lower confidence qualifier.
#
# RULE 3: HONEST JUSTIFICATION REQUIRED
# Every score must include a 1-2 sentence
# justification explaining what specific content
# or absence of content led to that score. The AI
# must never assign scores without explanation.
#
# RULE 4: NO FALSE PRECISION IN AGGREGATES
# When calculating averages or comparing sections,
# do not present results to decimal places that
# imply measurement precision. Use whole numbers
# or single decimals maximum. Frame comparisons
# as “relatively stronger/weaker” not “2.3 points
# better.”
#
# RULE 5: BATCH SIZE ENFORCEMENT
# Maximum 15 pages per batch (not 20 — quality
# degrades with larger batches). If user provides
# more than 15, ask them to split into smaller
# batches organized by section.
#
# RULE 6: CUMULATIVE INVENTORY MAINTENANCE
# The Site Inventory is a running document. Each
# batch ADDS rows — never overwrites previous
# entries. Present the full cumulative inventory
# after each batch so the user can track progress.
#
# RULE 7: CROSS-SECTION SIGNALS REQUIRE EVIDENCE
# When identifying overlaps, contradictions, or
# patterns across sections, cite specific content
# from both sections. Do not make vague claims
# like “there seems to be inconsistency.” Point
# to specific pages and content elements.
#
# RULE 8: WAIT FOR USER BETWEEN PHASES
# After each major deliverable (inventory update,
# scorecard, cross-section signals), STOP and
# wait for user confirmation before proceeding.
# The user may want to correct inferences, add
# context, or adjust scoring.
#
# RULE 9: OPPORTUNITIES MUST BE ACTIONABLE
# Every opportunity listed must include a specific
# action the user can take, not just a problem
# description. “CTA is weak” is a finding.
# “Add a specific benefit statement above the
# primary CTA button” is an opportunity.
#
# RULE 10: SERIES CONTEXT AWARENESS
# This recipe is part of the Website Analysis
# series. If the user mentions having run RCP-068
# (Website Navigation Auditor) previously, ask
# for their navigation findings to inform this
# analysis. If they haven’t, suggest starting
# with RCP-068 for structural overview first —
# but do not require it.
#
# RULE 11: NO INVENTED CONTENT
# The AI must never invent page content, user
# behavior data, or performance metrics. If the
# user has not provided content for a page, the
# AI assesses only what was provided. Gaps in
# information are noted, not filled with
# assumptions.
#
# =========================================================== WEBSITE_QUALITY_SCORECARD = Recipe(
recipe_id=”RCP-000-000-071″,
title=”Website Quality Scorecard”,
description=”Interactive website content audit that
builds a section-by-section inventory with
subjective quality assessments across five
dimensions. Produces comparative analysis and
prioritized improvement opportunities. Scores
reflect AI assessment of user-provided content,
not objective measurements.”,
category=”CAT-000-STANDALONE”,
subcategory=”SUBCAT-015-WEBSITE-ANALYSIS”,
difficulty=”medium”,
version=”2.00a-R”,
WPRM_FIELD_CRAFT_FLAVORS = “CRAFT Cowork”, parameters={
“website_name”: {
“type”: “string”,
“required”: True,
“description”: “Name of website to audit”,
“example”: “Acme Solutions Corporate Site”
},
“business_goal”: {
“type”: “string”,
“required”: True,
“description”: “Primary site objective”,
“example”: “Lead generation for B2B SaaS”,
“coaching”: “Be specific about what success
looks like. ‘Generate leads’ is too
vague. ‘Convert visitors into demo
requests for our project management
tool’ gives me calibration context.”
},
“target_audience”: {
“type”: “string”,
“required”: True,
“description”: “Who the site serves”,
“example”: “Mid-market IT directors
evaluating project management solutions”,
“coaching”: “Include role, company size, and
what they are looking for. This helps
me assess whether content speaks to
the right audience.”
},
“page_data”: {
“type”: “list”,
“required”: True,
“format”: “url, title, section, notes”,
“description”: “Page metadata for analysis”,
“example”: [
“acme.com/features, Features Overview,
Product, Main features landing page”,
“acme.com/pricing, Pricing Plans,
Product, Three-tier pricing model”,
“acme.com/about, About Us, Company,
Team and mission page”
],
“coaching”: “Organize pages by section.
Include content excerpts (copy-paste
key paragraphs) for more accurate
assessment. Without content, I can
only assess based on metadata — which
limits accuracy significantly.”
},
“custom_dimensions”: {
“type”: “list”,
“required”: False,
“default”: “none”,
“description”: “Extra scoring dimensions”,
“example”: “SEO Optimization, Mobile UX”,
“coaching”: “Only add dimensions you will
actually act on. Each added dimension
increases assessment time per page.”
}
}, prompt_template=” # ===========================================================
# WEBSITE QUALITY SCORECARD
# =========================================================== #H->AI::Directive: (Build interactive website content
inventory with section scorecards and comparative
analysis) # ———————————————————–
# PHASE 1: CONTEXT GATHERING AND CALIBRATION
# ———————————————————– #AI->H::Status: (Starting Website Quality Scorecard) Gather and confirm required context: WEBSITE: {website_name}
BUSINESS GOAL: {business_goal}
TARGET AUDIENCE: {target_audience} SCORING FRAMEWORK:
Five core dimensions, each rated 0-5: 1. CLARITY (0-5)
How clearly does this page communicate its
purpose and key messages to the target audience? 2. CONSISTENCY (0-5)
How well does this page align with messaging,
tone, and promises on other site sections? 3. DEPTH (0-5)
How comprehensively does this page address the
topic for its intended purpose? 4. CTA STRENGTH (0-5)
How effectively does this page guide visitors
toward a desired action? 5. TRUST SIGNALS (0-5)
How well does this page establish credibility
and reduce visitor uncertainty? IF {custom_dimensions} provided:
Add custom dimensions to the framework.
Define each with the same structure: name,
question it answers, 0-5 scale meaning. IMPORTANT FRAMING:
These scores reflect my subjective assessment of
the content you provide. They are structured
opinions, not measurements. Their value is in
creating a consistent framework for comparison
across your site sections — not as absolute
quality metrics. #AI->H::Question: (Does this scoring framework match
your assessment goals? Any dimensions to add,
remove, or redefine before we begin?) [WAIT — Do not proceed until user confirms framework] # ———————————————————–
# PHASE 2: BATCH INTAKE AND VALIDATION
# ———————————————————– Upon receiving page data batch: BATCH VALIDATION
================
Pages Received: [N]
Section: [section name if uniform] IF batch exceeds 15 pages:
#AI->H::Caution: (Batch contains [N] pages.
For assessment quality, I recommend
splitting into batches of 15 or fewer.
Shall I proceed with the first 15, or
would you like to reorganize?)
[WAIT for user decision] VALIDATION CHECKS:
Duplicates: [count or ‘None found’]
Missing Titles: [count or ‘All present’]
Content Provided: [Yes — full excerpts /
Partial — some pages / No — metadata only] IF no content excerpts provided:
#AI->H::Caution: (No content excerpts included
in this batch. Scoring will be based on
metadata only, which significantly limits
accuracy. All scores will be marked
[metadata-only]. For better results,
paste key paragraphs from each page.) Proceed with validated pages. [WAIT — Confirm validation results before scoring] # ———————————————————–
# PHASE 3: SITE INVENTORY UPDATE
# ———————————————————– For each validated page, add to cumulative inventory: SITE INVENTORY — BATCH [N] ADDITIONS
=============================================
| Sect | URL | Title | Type | Purpose |
| | | | | |
| [s] | [url] | [t] | [inf] | [purpose] |
| [s] | [url] | [t] | [inf] | [purpose] |
============================================= Content Type (inferred from provided content):
Informational / Transactional /
Trust-Building / Resource / Landing / Support Key Purpose (inferred):
One sentence describing page objective. Page-Level Observations (per page):
Issues (up to 3): Specific problems found
Opportunities (up to 3): Specific actions
to improve (per RULE 9 — actionable) IF previous batches exist:
Show running total: ‘Cumulative inventory:
[N] pages across [N] sections’ Present inventory additions to user. [WAIT — User reviews inferences for accuracy] # ———————————————————–
# PHASE 4: SECTION SCORECARD
# ———————————————————– After completing all pages in a section: SECTION SCORECARD: [Section Name]
=========================================
Pages Assessed: [N]
Content Basis: [Full excerpts / Metadata only] DIMENSION SCORE JUSTIFICATION
—————————————–
Clarity [0-5] [1-2 sentence reason
citing specific content]
Consistency [0-5] [1-2 sentence reason]
Depth [0-5] [1-2 sentence reason]
CTA Strength [0-5 [1-2 sentence reason]
or N/A]
Trust Signals [0-5] [1-2 sentence reason]
—————————————–
SECTION TOTAL [X]/[MAX] N/A HANDLING: If a dimension is scored N/A
(legitimate for CTA Strength on pages that
do not require a call to action, such as
About or legal pages), reduce the section
maximum accordingly. A section with one N/A
dimension scores out of 20, not 25. Score
Context thresholds scale proportionally. Score Context (per scored dimension):
80-100%: Strong — minor refinements only
60-79%: Solid — targeted improvements
40-59%: Mixed — systematic work needed
0-39%: Weak — significant gaps present Note: Scores reflect AI assessment of provided
content and may differ from user experience
testing or analytics data. IF metadata-only:
All scores marked [metadata-only].
Confidence: LOW. Provide content excerpts
for a more accurate assessment. Present scorecard to user. [WAIT — User reviews scores and may challenge
or provide additional context] # ———————————————————–
# PHASE 5: CROSS-SECTION SIGNALS
# ———————————————————– After each section scorecard (starting from section 2): CROSS-SECTION SIGNALS
===================== OVERLAPS:
[Specific content on Page X in Section A
covers similar ground as Page Y in Section B.
Cite what both pages say.] CONTRADICTIONS:
[Page X says ‘[specific claim]’ but Page Y
says ‘[conflicting claim].’ This may confuse
visitors moving between sections.] EMERGING PATTERNS:
[Across [N] sections analyzed, [pattern
observed] — e.g., ‘CTA language is
consistently generic across all sections,
suggesting a site-wide messaging gap.’] IF first section analyzed:
‘This is the first section scored. Cross-
section patterns will emerge as more sections
are processed.’ RUNNING COMPARISON (if 2+ sections):
| Section | Score | Strongest Dim | Weakest |
|———–|——-|—————|———–|
| [Sect A] | [X] | [Dim] | [Dim] |
| [Sect B] | [X] | [Dim] | [Dim] | #AI->H::Question: (Section [Name] complete. Ready
for next batch? Please provide pages from your
next section.) [WAIT — User provides next batch or requests final] # ———————————————————–
# PHASE 6: FINAL COMPARATIVE ANALYSIS
# ———————————————————– After user indicates all sections are complete: COMPLETE SITE INVENTORY: {website_name}
=========================================
[Full cumulative inventory table — all sections]
=========================================
Total Pages Assessed: [N]
Total Sections: [N]
Content Basis: [Full/Partial/Metadata-only] SECTION COMPARISON
==================
| Section | Score | Rank | Strongest | Weakest |
|———–|——–|——|————-|————|
| [Highest] | [X]% | 1 | [Dimension] | [Dimension]|
| [Next] | [X]% | 2 | [Dimension] | [Dimension]|
| [Lowest] | [X]% | [N] | [Dimension] | [Dimension]|
(Percentages normalize for N/A dimensions) DIMENSION ANALYSIS (across all sections)
==================
Strongest: [Dimension] — Average [X]/5
Weakest: [Dimension] — Average [X]/5 TOP 10 IMPROVEMENT OPPORTUNITIES
(ranked by estimated impact)
================================
1. [Page]: [Specific actionable recommendation]
Impact: [High/Medium/Low]
Effort: [High/Medium/Low]
Section Score Affected: [Dimension] 2. [Page]: [Specific actionable recommendation]
Impact: [High/Medium/Low]
Effort: [High/Medium/Low] [Continue to 10 or total available] SITE-WIDE OBSERVATIONS
======================
[2-3 paragraphs synthesizing patterns across
all sections — what the site does well overall,
where systemic gaps exist, and what single
improvement would have the widest impact] METHODOLOGY NOTE:
This scorecard reflects structured subjective
assessment based on the content you provided.
Scores enable relative comparison across your
site sections but should be validated against
actual user behavior data and conversion
analytics where available. #AI->H::Delivery: (Website Quality Scorecard
complete for {website_name}. [N] pages across
[N] sections assessed.)
“, outputs={
“SITE_INVENTORY”: {
“format”: “Cumulative table, exportable”,
“purpose”: “Living document for tracking”
},
“SECTION_SCORECARDS”: {
“format”: “0-5 subjective ratings with
justification per dimension”,
“purpose”: “Structured quality assessment”
},
“CROSS_SECTION_SIGNALS”: {
“format”: “Evidence-cited pattern bullets”,
“purpose”: “Systemic issue detection”
},
“COMPARATIVE_ANALYSIS”: {
“format”: “Rankings, dimension analysis,
prioritized opportunities”,
“purpose”: “Decision-making support”
}
}, integration_notes=”
SERIES: Website Analysis (2 of 3)
1. RCP-068 Website Navigation Auditor (structural)
2. RCP-071 Website Quality Scorecard (this recipe)
3. RCP-072 Website Evidence Auditor (strategic) RECOMMENDED WORKFLOW:
1. Run RCP-068 for structural navigation overview
2. Use 068 findings to prioritize sections for 071
3. Build inventory and scorecards with 071
4. Deep-dive critical sections with 072
“
)
# WEBSITE QUALITY SCORECARD
# Recipe ID: RCP-000-000-071
# Version: v2.00a-R (QA Revised)
# Series: Website Analysis (2 of 3)
# =========================================================== # ===========================================================
# BEHAVIORAL RULES
# ===========================================================
#
# RULE 1: SCORES ARE SUBJECTIVE ASSESSMENTS
# All 0-5 scores represent the AI’s subjective
# assessment based on available information, NOT
# objective measurements. Every scorecard must
# include a brief note: “Scores reflect AI
# assessment of provided content and may differ
# from user experience testing or analytics data.”
#
# RULE 2: NO SCORING WITHOUT CONTENT
# If the user provides only URLs and titles without
# content excerpts, the AI must flag that scoring
# accuracy is severely limited. Scores based on
# metadata only must be marked “[metadata-only]”
# and carry a lower confidence qualifier.
#
# RULE 3: HONEST JUSTIFICATION REQUIRED
# Every score must include a 1-2 sentence
# justification explaining what specific content
# or absence of content led to that score. The AI
# must never assign scores without explanation.
#
# RULE 4: NO FALSE PRECISION IN AGGREGATES
# When calculating averages or comparing sections,
# do not present results to decimal places that
# imply measurement precision. Use whole numbers
# or single decimals maximum. Frame comparisons
# as “relatively stronger/weaker” not “2.3 points
# better.”
#
# RULE 5: BATCH SIZE ENFORCEMENT
# Maximum 15 pages per batch (not 20 — quality
# degrades with larger batches). If user provides
# more than 15, ask them to split into smaller
# batches organized by section.
#
# RULE 6: CUMULATIVE INVENTORY MAINTENANCE
# The Site Inventory is a running document. Each
# batch ADDS rows — never overwrites previous
# entries. Present the full cumulative inventory
# after each batch so the user can track progress.
#
# RULE 7: CROSS-SECTION SIGNALS REQUIRE EVIDENCE
# When identifying overlaps, contradictions, or
# patterns across sections, cite specific content
# from both sections. Do not make vague claims
# like “there seems to be inconsistency.” Point
# to specific pages and content elements.
#
# RULE 8: WAIT FOR USER BETWEEN PHASES
# After each major deliverable (inventory update,
# scorecard, cross-section signals), STOP and
# wait for user confirmation before proceeding.
# The user may want to correct inferences, add
# context, or adjust scoring.
#
# RULE 9: OPPORTUNITIES MUST BE ACTIONABLE
# Every opportunity listed must include a specific
# action the user can take, not just a problem
# description. “CTA is weak” is a finding.
# “Add a specific benefit statement above the
# primary CTA button” is an opportunity.
#
# RULE 10: SERIES CONTEXT AWARENESS
# This recipe is part of the Website Analysis
# series. If the user mentions having run RCP-068
# (Website Navigation Auditor) previously, ask
# for their navigation findings to inform this
# analysis. If they haven’t, suggest starting
# with RCP-068 for structural overview first —
# but do not require it.
#
# RULE 11: NO INVENTED CONTENT
# The AI must never invent page content, user
# behavior data, or performance metrics. If the
# user has not provided content for a page, the
# AI assesses only what was provided. Gaps in
# information are noted, not filled with
# assumptions.
#
# =========================================================== WEBSITE_QUALITY_SCORECARD = Recipe(
recipe_id=”RCP-000-000-071″,
title=”Website Quality Scorecard”,
description=”Interactive website content audit that
builds a section-by-section inventory with
subjective quality assessments across five
dimensions. Produces comparative analysis and
prioritized improvement opportunities. Scores
reflect AI assessment of user-provided content,
not objective measurements.”,
category=”CAT-000-STANDALONE”,
subcategory=”SUBCAT-015-WEBSITE-ANALYSIS”,
difficulty=”medium”,
version=”2.00a-R”,
WPRM_FIELD_CRAFT_FLAVORS = “CRAFT Cowork”, parameters={
“website_name”: {
“type”: “string”,
“required”: True,
“description”: “Name of website to audit”,
“example”: “Acme Solutions Corporate Site”
},
“business_goal”: {
“type”: “string”,
“required”: True,
“description”: “Primary site objective”,
“example”: “Lead generation for B2B SaaS”,
“coaching”: “Be specific about what success
looks like. ‘Generate leads’ is too
vague. ‘Convert visitors into demo
requests for our project management
tool’ gives me calibration context.”
},
“target_audience”: {
“type”: “string”,
“required”: True,
“description”: “Who the site serves”,
“example”: “Mid-market IT directors
evaluating project management solutions”,
“coaching”: “Include role, company size, and
what they are looking for. This helps
me assess whether content speaks to
the right audience.”
},
“page_data”: {
“type”: “list”,
“required”: True,
“format”: “url, title, section, notes”,
“description”: “Page metadata for analysis”,
“example”: [
“acme.com/features, Features Overview,
Product, Main features landing page”,
“acme.com/pricing, Pricing Plans,
Product, Three-tier pricing model”,
“acme.com/about, About Us, Company,
Team and mission page”
],
“coaching”: “Organize pages by section.
Include content excerpts (copy-paste
key paragraphs) for more accurate
assessment. Without content, I can
only assess based on metadata — which
limits accuracy significantly.”
},
“custom_dimensions”: {
“type”: “list”,
“required”: False,
“default”: “none”,
“description”: “Extra scoring dimensions”,
“example”: “SEO Optimization, Mobile UX”,
“coaching”: “Only add dimensions you will
actually act on. Each added dimension
increases assessment time per page.”
}
}, prompt_template=” # ===========================================================
# WEBSITE QUALITY SCORECARD
# =========================================================== #H->AI::Directive: (Build interactive website content
inventory with section scorecards and comparative
analysis) # ———————————————————–
# PHASE 1: CONTEXT GATHERING AND CALIBRATION
# ———————————————————– #AI->H::Status: (Starting Website Quality Scorecard) Gather and confirm required context: WEBSITE: {website_name}
BUSINESS GOAL: {business_goal}
TARGET AUDIENCE: {target_audience} SCORING FRAMEWORK:
Five core dimensions, each rated 0-5: 1. CLARITY (0-5)
How clearly does this page communicate its
purpose and key messages to the target audience? 2. CONSISTENCY (0-5)
How well does this page align with messaging,
tone, and promises on other site sections? 3. DEPTH (0-5)
How comprehensively does this page address the
topic for its intended purpose? 4. CTA STRENGTH (0-5)
How effectively does this page guide visitors
toward a desired action? 5. TRUST SIGNALS (0-5)
How well does this page establish credibility
and reduce visitor uncertainty? IF {custom_dimensions} provided:
Add custom dimensions to the framework.
Define each with the same structure: name,
question it answers, 0-5 scale meaning. IMPORTANT FRAMING:
These scores reflect my subjective assessment of
the content you provide. They are structured
opinions, not measurements. Their value is in
creating a consistent framework for comparison
across your site sections — not as absolute
quality metrics. #AI->H::Question: (Does this scoring framework match
your assessment goals? Any dimensions to add,
remove, or redefine before we begin?) [WAIT — Do not proceed until user confirms framework] # ———————————————————–
# PHASE 2: BATCH INTAKE AND VALIDATION
# ———————————————————– Upon receiving page data batch: BATCH VALIDATION
================
Pages Received: [N]
Section: [section name if uniform] IF batch exceeds 15 pages:
#AI->H::Caution: (Batch contains [N] pages.
For assessment quality, I recommend
splitting into batches of 15 or fewer.
Shall I proceed with the first 15, or
would you like to reorganize?)
[WAIT for user decision] VALIDATION CHECKS:
Duplicates: [count or ‘None found’]
Missing Titles: [count or ‘All present’]
Content Provided: [Yes — full excerpts /
Partial — some pages / No — metadata only] IF no content excerpts provided:
#AI->H::Caution: (No content excerpts included
in this batch. Scoring will be based on
metadata only, which significantly limits
accuracy. All scores will be marked
[metadata-only]. For better results,
paste key paragraphs from each page.) Proceed with validated pages. [WAIT — Confirm validation results before scoring] # ———————————————————–
# PHASE 3: SITE INVENTORY UPDATE
# ———————————————————– For each validated page, add to cumulative inventory: SITE INVENTORY — BATCH [N] ADDITIONS
=============================================
| Sect | URL | Title | Type | Purpose |
| | | | | |
| [s] | [url] | [t] | [inf] | [purpose] |
| [s] | [url] | [t] | [inf] | [purpose] |
============================================= Content Type (inferred from provided content):
Informational / Transactional /
Trust-Building / Resource / Landing / Support Key Purpose (inferred):
One sentence describing page objective. Page-Level Observations (per page):
Issues (up to 3): Specific problems found
Opportunities (up to 3): Specific actions
to improve (per RULE 9 — actionable) IF previous batches exist:
Show running total: ‘Cumulative inventory:
[N] pages across [N] sections’ Present inventory additions to user. [WAIT — User reviews inferences for accuracy] # ———————————————————–
# PHASE 4: SECTION SCORECARD
# ———————————————————– After completing all pages in a section: SECTION SCORECARD: [Section Name]
=========================================
Pages Assessed: [N]
Content Basis: [Full excerpts / Metadata only] DIMENSION SCORE JUSTIFICATION
—————————————–
Clarity [0-5] [1-2 sentence reason
citing specific content]
Consistency [0-5] [1-2 sentence reason]
Depth [0-5] [1-2 sentence reason]
CTA Strength [0-5 [1-2 sentence reason]
or N/A]
Trust Signals [0-5] [1-2 sentence reason]
—————————————–
SECTION TOTAL [X]/[MAX] N/A HANDLING: If a dimension is scored N/A
(legitimate for CTA Strength on pages that
do not require a call to action, such as
About or legal pages), reduce the section
maximum accordingly. A section with one N/A
dimension scores out of 20, not 25. Score
Context thresholds scale proportionally. Score Context (per scored dimension):
80-100%: Strong — minor refinements only
60-79%: Solid — targeted improvements
40-59%: Mixed — systematic work needed
0-39%: Weak — significant gaps present Note: Scores reflect AI assessment of provided
content and may differ from user experience
testing or analytics data. IF metadata-only:
All scores marked [metadata-only].
Confidence: LOW. Provide content excerpts
for a more accurate assessment. Present scorecard to user. [WAIT — User reviews scores and may challenge
or provide additional context] # ———————————————————–
# PHASE 5: CROSS-SECTION SIGNALS
# ———————————————————– After each section scorecard (starting from section 2): CROSS-SECTION SIGNALS
===================== OVERLAPS:
[Specific content on Page X in Section A
covers similar ground as Page Y in Section B.
Cite what both pages say.] CONTRADICTIONS:
[Page X says ‘[specific claim]’ but Page Y
says ‘[conflicting claim].’ This may confuse
visitors moving between sections.] EMERGING PATTERNS:
[Across [N] sections analyzed, [pattern
observed] — e.g., ‘CTA language is
consistently generic across all sections,
suggesting a site-wide messaging gap.’] IF first section analyzed:
‘This is the first section scored. Cross-
section patterns will emerge as more sections
are processed.’ RUNNING COMPARISON (if 2+ sections):
| Section | Score | Strongest Dim | Weakest |
|———–|——-|—————|———–|
| [Sect A] | [X] | [Dim] | [Dim] |
| [Sect B] | [X] | [Dim] | [Dim] | #AI->H::Question: (Section [Name] complete. Ready
for next batch? Please provide pages from your
next section.) [WAIT — User provides next batch or requests final] # ———————————————————–
# PHASE 6: FINAL COMPARATIVE ANALYSIS
# ———————————————————– After user indicates all sections are complete: COMPLETE SITE INVENTORY: {website_name}
=========================================
[Full cumulative inventory table — all sections]
=========================================
Total Pages Assessed: [N]
Total Sections: [N]
Content Basis: [Full/Partial/Metadata-only] SECTION COMPARISON
==================
| Section | Score | Rank | Strongest | Weakest |
|———–|——–|——|————-|————|
| [Highest] | [X]% | 1 | [Dimension] | [Dimension]|
| [Next] | [X]% | 2 | [Dimension] | [Dimension]|
| [Lowest] | [X]% | [N] | [Dimension] | [Dimension]|
(Percentages normalize for N/A dimensions) DIMENSION ANALYSIS (across all sections)
==================
Strongest: [Dimension] — Average [X]/5
Weakest: [Dimension] — Average [X]/5 TOP 10 IMPROVEMENT OPPORTUNITIES
(ranked by estimated impact)
================================
1. [Page]: [Specific actionable recommendation]
Impact: [High/Medium/Low]
Effort: [High/Medium/Low]
Section Score Affected: [Dimension] 2. [Page]: [Specific actionable recommendation]
Impact: [High/Medium/Low]
Effort: [High/Medium/Low] [Continue to 10 or total available] SITE-WIDE OBSERVATIONS
======================
[2-3 paragraphs synthesizing patterns across
all sections — what the site does well overall,
where systemic gaps exist, and what single
improvement would have the widest impact] METHODOLOGY NOTE:
This scorecard reflects structured subjective
assessment based on the content you provided.
Scores enable relative comparison across your
site sections but should be validated against
actual user behavior data and conversion
analytics where available. #AI->H::Delivery: (Website Quality Scorecard
complete for {website_name}. [N] pages across
[N] sections assessed.)
“, outputs={
“SITE_INVENTORY”: {
“format”: “Cumulative table, exportable”,
“purpose”: “Living document for tracking”
},
“SECTION_SCORECARDS”: {
“format”: “0-5 subjective ratings with
justification per dimension”,
“purpose”: “Structured quality assessment”
},
“CROSS_SECTION_SIGNALS”: {
“format”: “Evidence-cited pattern bullets”,
“purpose”: “Systemic issue detection”
},
“COMPARATIVE_ANALYSIS”: {
“format”: “Rankings, dimension analysis,
prioritized opportunities”,
“purpose”: “Decision-making support”
}
}, integration_notes=”
SERIES: Website Analysis (2 of 3)
1. RCP-068 Website Navigation Auditor (structural)
2. RCP-071 Website Quality Scorecard (this recipe)
3. RCP-072 Website Evidence Auditor (strategic) RECOMMENDED WORKFLOW:
1. Run RCP-068 for structural navigation overview
2. Use 068 findings to prioritize sections for 071
3. Build inventory and scorecards with 071
4. Deep-dive critical sections with 072
“
)
{
“schema”: “CRAFT_AI_TO_AI_COMMUNICATION”,
“schema_version”: “1.1”,
“profile”: “standalone-recipe”,
“recipe_id”: “RCP-000-000-071”,
“title”: “Website Quality Scorecard”,
“version”: “2.00a-R”,
“status”: “Phase 3 LOCKED”,
“category”: “CAT-000-STANDALONE”,
“subcategory”: “SUBCAT-015-WEBSITE-ANALYSIS”,
“difficulty”: “medium”,
“series”: {
“name”: “Website Analysis”,
“position”: “2 of 3”,
“companions”: [“RCP-068”, “RCP-072”]
},
“family”: {
“name”: “Website Analysis”,
“position”: “2 of 3”,
“type”: “series”
},
“description”: “Interactive website content audit that builds a section-by-section inventory with subjective quality assessments across five dimensions. Produces comparative analysis and prioritized improvement opportunities. Scores reflect AI assessment of user-provided content, not objective measurements.”,
“parameters”: [
{
“name”: “website_name”,
“type”: “string”,
“required”: true,
“description”: “Name of website to audit”
},
{
“name”: “business_goal”,
“type”: “string”,
“required”: true,
“description”: “Primary site objective”
},
{
“name”: “target_audience”,
“type”: “string”,
“required”: true,
“description”: “Who the site serves”
},
{
“name”: “page_data”,
“type”: “list”,
“required”: true,
“format”: “url, title, section, notes”,
“description”: “Page metadata for analysis”
},
{
“name”: “custom_dimensions”,
“type”: “list”,
“required”: false,
“description”: “Extra scoring dimensions beyond the five core”
}
],
“interactive_inputs”: [
{
“name”: “batch_data”,
“type”: “multi-turn”,
“description”: “User provides page data in section-organized batches across multiple turns; AI validates, scores, and builds cumulative inventory per batch”
}
],
“behavioral_rules”: [
“R1: Scores are subjective assessments, not objective measurements”,
“R2: No scoring without content — metadata-only flagged with qualifier”,
“R3: Honest justification required for every score”,
“R4: No false precision in aggregates — single decimal max”,
“R5: Batch size enforcement — 15 pages maximum”,
“R6: Cumulative inventory maintenance — batches add, never overwrite”,
“R7: Cross-section signals require specific evidence citations”,
“R8: Wait for user between phases — 4 WAIT gates”,
“R9: Opportunities must be actionable, not just problem descriptions”,
“R10: Series context awareness — leverage RCP-068 findings if available”,
“R11: No invented content — gaps noted, not filled with assumptions”
],
“delivery_structure”: {
“type”: “interactive-multi-turn”,
“phases”: 6,
“wait_gates”: 4,
“phase_list”: [
“Phase 1: Context Gathering and Calibration (WAIT)”,
“Phase 2: Batch Intake and Validation (WAIT)”,
“Phase 3: Site Inventory Update (WAIT)”,
“Phase 4: Section Scorecard (WAIT)”,
“Phase 5: Cross-Section Signals (repeats per section)”,
“Phase 6: Final Comparative Analysis”
],
“scoring_framework”: {
“dimensions”: [“Clarity”, “Consistency”, “Depth”, “CTA Strength”, “Trust Signals”],
“scale”: “0-5 per dimension, N/A supported for CTA Strength”,
“aggregation”: “percentage-based Score Context thresholds to normalize for N/A dimensions”
}
},
“outputs”: [
“SITE_INVENTORY: Cumulative table across batches”,
“SECTION_SCORECARDS: 0-5 subjective ratings with justification per dimension”,
“CROSS_SECTION_SIGNALS: Evidence-cited pattern analysis”,
“COMPARATIVE_ANALYSIS: Rankings, dimension analysis, top 10 prioritized opportunities”
],
“relationships”: {
“series_companions”: [
{
“recipe_id”: “RCP-000-000-068”,
“title”: “Website Navigation Auditor”,
“relationship”: “series-precursor”,
“note”: “Structural navigation audit — run first for site overview”
},
{
“recipe_id”: “RCP-000-000-072”,
“title”: “Website Evidence Auditor”,
“relationship”: “series-successor”,
“note”: “Deep evidence-based audit — use scorecard results to target which sections need deeper analysis”
}
]
},
“lessons_learned”: [
{
“id”: “LL-071-001”,
“category”: “STRUCTURAL”,
“summary”: “Scoring frameworks with optional dimensions need N/A handling and proportional thresholds”,
“detail”: “Fixed-denominator scoring (/25 assuming 5 dimensions) breaks when a dimension is legitimately not applicable. Percentage-based Score Context thresholds (80-100% Strong, 60-79% Solid, etc.) normalize cleanly regardless of how many dimensions are scored. Applied to CTA Strength N/A case.”
},
{
“id”: “LL-071-002”,
“category”: “CROSS-DOCUMENT”,
“summary”: “Supplementary guidance must not introduce behavior the recipe template cannot accommodate”,
“detail”: “Extended AI Guidance introduced N/A scoring for CTA Strength but the recipe template only supported 0-5 with fixed /25 totals. Cross-document consistency requires the recipe to structurally support any behavior the guidance recommends.”
}
],
“pipeline_metadata”: {
“pipeline_run”: 45,
“crpw_number”: 44,
“type”: “Standalone”,
“intake_session”: “H086”,
“evaluation_findings”: 1,
“finding_disposition”: “ACCEPT 1/1”,
“rec_19_applied”: true,
“rec_19_detail”: “CRAFT_FLAVORS ABSENT — 10th consecutive, 3rd ADD”
}
}
Show/Hide accordion — “Extended Information for the AI” section (AI-to-AI execution guidance, failure modes, tone calibration, common mistakes)
