| name | agency-report-pdf |
| description | Unified PDF report generator — combines all audit scores into a professional client-ready PDF |
Unified Agency PDF Report Generator
You are the PDF Report Generator for the AI Agency Command Center. When the user runs /agency report-pdf, you scan the current directory for all audit output files, extract scores and findings from each available audit, prepare a structured JSON data file, and run the Python PDF generation script to produce a professional, multi-page AGENCY-REPORT.pdf.
Trigger
This skill activates when the user runs:
/agency report-pdf
No arguments required. This command operates on whatever audit files exist in the current working directory.
Overview of the PDF Generation Pipeline
[Scan Directory] → [Extract Data from Audit Files] → [Build JSON Structure] → [Write agency_data.json] → [Run Python Script] → [AGENCY-REPORT.pdf]
The Python script at ~/.claude/skills/agency/scripts/generate_agency_pdf.py handles all PDF rendering. Your job is to prepare the data. The script expects a file called agency_data.json in the current working directory.
Step 1 — Scan for Available Audit Files
Search the current working directory for all audit output files using Glob. Check for each of these file patterns:
Agency-Level Files
AGENCY-ONBOARD-*.md → Primary source for composite scores
AGENCY-PROPOSAL-*.md → Proposal data for service recommendations
Individual Tool Suite Files
MARKETING-AUDIT*.md → Marketing score and findings
REPUTATION-AUDIT-*.md → Reputation score and findings
GEO-AUDIT-*.md → GEO/SEO score and findings
LEGAL-COMPLIANCE-*.md → Legal score and findings
PROSPECT-ANALYSIS*.md → Sales/opportunity score and findings
SALES-RESEARCH*.md → Additional sales data
Supplementary Files (for enrichment)
REPUTATION-REVIEWS*.md → Review data for reputation section
REPUTATION-SENTIMENT*.md → Sentiment data
GEO-CITABILITY*.md → Citability details
GEO-SCHEMA*.md → Schema markup details
GEO-CRAWLERS*.md → Crawler access data
MARKETING-SEO*.md → SEO detail data
MARKETING-FUNNEL*.md → Funnel data
LEGAL-PRIVACY*.md → Privacy policy details
LEGAL-TERMS*.md → Terms of service details
If NO audit files are found at all, display an error:
No audit files found in the current directory.
Run /agency onboard <url> first to generate audit data, then try again.
Step 2 — Extract Data from Each Audit File
Read each discovered file and extract the relevant data points. Use careful parsing — scores may appear in different formats across files.
2A — Extract from Agency Onboard Report (AGENCY-ONBOARD-*.md)
This is the richest data source. If present, it contains everything. Look for:
- Company name — Usually in the title or first heading
- Agency Score — Look for patterns like "Agency Score: XX/100", "Composite Score: XX", or a score table
- Agency Grade — Look for "Grade: X" or grade in the score table
- Individual scores — Look for a score breakdown table or section with:
- Marketing Score (or Marketing: XX/100)
- Reputation Score
- GEO Score (or GEO/SEO Score)
- Legal Score
- Sales Score (or Opportunity Score)
- Critical findings — Look for sections titled "Critical Findings", "Key Issues", or "Problems Found". Extract the top 3 from each team.
- Quick wins — Look for sections titled "Quick Wins", "Easy Fixes", or "Low-Hanging Fruit". Extract the top 3 from each team.
- Recommended service tier — Look for "Recommended", "Service Package", "Pricing", or tier names (Essentials, Growth, Full Agency)
- 90-day action plan — Look for phased roadmap, timeline, or action plan sections
- Company profile data — Industry, location, business type, website URL
2B — Extract from Individual Marketing Audit (MARKETING-AUDIT*.md)
If no agency onboard exists, or to supplement it:
- Marketing Score — Look for "Marketing Score: XX/100", "Overall Score: XX", or similar
- Copy quality assessment — Rating or description of website copy
- SEO status — Meta tags, headings, content structure assessment
- Conversion elements — CTAs, forms, social proof evaluation
- Content strategy — Blog presence, thought leadership assessment
- Critical findings — Top 3 marketing issues
- Quick wins — Top 3 easy marketing fixes
- Recommended marketing services — With pricing if available
2C — Extract from Reputation Audit (REPUTATION-AUDIT-*.md)
- Reputation Score — Look for "Reputation Score: XX/100" or similar
- Google rating — Star rating (e.g., 3.8/5.0)
- Review count — Total number of Google reviews
- Sentiment breakdown — Positive/negative/neutral percentages
- Response rate — Percentage of negative reviews with owner responses
- Competitor comparison — How this business compares to local competitors
- Critical findings — Top 3 reputation issues
- Quick wins — Top 3 easy reputation fixes
2D — Extract from GEO Audit (GEO-AUDIT-*.md)
- GEO Score — Look for "GEO Score: XX/100" or "AI Visibility Score"
- Citability Score — How likely AI systems cite this content
- AI crawler access — Which AI crawlers are allowed/blocked
- Schema markup status — Present, partial, or missing
- Platform readiness — Scores for ChatGPT, Perplexity, Gemini, Google AI Overviews
- Critical findings — Top 3 GEO/SEO issues
- Quick wins — Top 3 easy GEO fixes
2E — Extract from Legal Compliance (LEGAL-COMPLIANCE-*.md)
- Legal Score — Look for "Legal Score: XX/100" or "Compliance Score"
- Privacy policy status — Present/missing, compliant/non-compliant
- Terms of service status — Present/missing, issues found
- Cookie consent — Compliant/non-compliant
- ADA/accessibility — Status and issues
- Critical findings — Top 3 compliance gaps
- Quick wins — Top 3 easy compliance fixes
2F — Extract from Sales/Prospect Analysis (PROSPECT-ANALYSIS*.md)
- Sales Score — Look for "Opportunity Score: XX/100" or "Sales Score"
- Company size — Employee count, revenue estimates
- Industry — Business category
- Decision makers — Names, titles, contact strategies
- Budget capacity — Estimated budget
- Critical findings — Top 3 sales insights
- Quick wins — Top 3 engagement opportunities
Step 3 — Calculate Composite Scores (if not already available)
If the agency onboard file is present and has a composite score, use it directly.
If individual scores exist but no composite, calculate:
Agency Score = (Marketing x 0.25) + (Reputation x 0.20) + (GEO x 0.20) + (Legal x 0.15) + (Sales x 0.20)
If some scores are missing, recalculate weights proportionally across available scores. For example, if only Marketing (25%), Reputation (20%), and GEO (20%) are available:
Total available weight = 0.25 + 0.20 + 0.20 = 0.65
Adjusted: Marketing = 0.25/0.65, Reputation = 0.20/0.65, GEO = 0.20/0.65
Grade Assignment
| Score | Grade |
|---|
| 85-100 | A+ |
| 70-84 | A |
| 55-69 | B |
| 40-54 | C |
| 25-39 | D |
| 0-24 | F |
Step 4 — Determine Service Tier Recommendation
Based on the composite score and number of critical findings:
Tier 1 — Essentials ($500-$1,500/month)
- Agency Score 55+ (Grade B or better)
- Fewer than 8 critical findings total
- Focus: monitoring, basic fixes, maintenance
Tier 2 — Growth ($1,500-$3,500/month)
- Agency Score 35-54 (Grade C-D)
- 8-15 critical findings total
- Focus: active improvement across multiple dimensions
Tier 3 — Full Agency ($3,500-$7,500/month)
- Agency Score below 35 (Grade D-F)
- 15+ critical findings total
- Focus: complete overhaul and ongoing management
If a proposal file exists, use the pricing from the proposal instead of estimating.
Step 5 — Build the JSON Data Structure
Construct the following JSON structure. All fields are required. Use null for unavailable data, never omit keys.
{
"company_name": "Business Name",
"date": "2026-04-05",
"website_url": "https://example.com",
"industry": "Industry category",
"location": "City, State",
"agency_score": 52,
"agency_grade": "C",
"marketing_score": 45,
"reputation_score": 62,
"geo_score": 38,
"legal_score": 55,
"sales_score": 68,
"scores_available": {
"marketing": true,
"reputation"
Step 6 — Write the JSON File
Write the constructed JSON to agency_data.json in the current working directory:
Use the Write tool to create agency_data.json with the full JSON structure
Validate the JSON is well-formed before writing. Ensure:
- All scores are integers 0-100 or null
- All arrays have at most 4 items (to fit PDF layout)
- All strings are properly escaped
- The date is in YYYY-MM-DD format
- No trailing commas
Step 7 — Run the PDF Generation Script
Execute the Python PDF generator:
python3 ~/.claude/skills/agency/scripts/generate_agency_pdf.py
The script reads agency_data.json from the current directory and outputs AGENCY-REPORT.pdf to the current directory.
If the Script Fails
-
Script not found — Inform the user:
PDF generation script not found at ~/.claude/skills/agency/scripts/generate_agency_pdf.py
The agency_data.json has been prepared. You can generate the PDF once the script is installed.
-
Python dependency missing — The script requires reportlab. If the import fails:
pip3 install reportlab
Then retry the script.
-
JSON parsing error — Re-validate the JSON structure. Common issues:
- Unescaped quotes in finding text
- Null values where strings are expected
- Missing required fields
-
Other errors — Display the full error output and suggest the user check the script.
Step 8 — Confirm Output
After successful PDF generation, display:
================================================================
AGENCY REPORT PDF GENERATED
================================================================
File: AGENCY-REPORT.pdf
Client: [Company Name]
Date: [Date]
Score: [Agency Score]/100 (Grade [Grade])
Pages: [Estimated page count based on data]
Scores included:
Marketing: [score or "N/A"]
Reputation: [score or "N/A"]
GEO/SEO: [score or "N/A"]
Legal: [score or "N/A"]
Sales: [score or "N/A"]
Data source: agency_data.json
The PDF has been saved to the current directory.
Share it with your client as a professional audit summary.
================================================================
Handling Partial Data
Not all 5 audits need to be present. The report adapts to whatever data is available:
- Only 1 audit available — Generate a single-dimension report. Note which audits are missing and recommend running them.
- 2-4 audits available — Generate a partial composite score using proportional weights. Clearly mark which dimensions were not assessed.
- All 5 audits available — Full comprehensive report.
For missing dimensions, the JSON should use null for the score and empty arrays for findings:
{
"legal_score": null,
"legal_findings": {
"critical": [],
"quick_wins": [],
"summary": "Legal compliance audit not yet performed."
}
}
Data Quality Rules
- Never fabricate scores — Only include scores actually found in audit files. Use null for missing data.
- Preserve original wording — Copy findings verbatim from audit files. Do not rephrase or embellish.
- Trim to fit — Each findings array should have exactly 3-4 items max. If the audit has more, pick the highest-impact ones.
- Validate score ranges — Scores must be 0-100 integers. If a file has a score outside this range, cap it.
- Date accuracy — Use the date from the most recent audit file, not today's date, unless today is the audit date.
Multiple Clients in Directory
If the current directory contains audit files for multiple businesses:
- Identify all unique business names from file names
- Ask the user which client the report should be for
- Filter to only that client's files
- If the user says "all" — generate one report for the most recently audited client and note others are available
Do NOT silently merge data from different businesses into one report.
