geo-report-pdf – skill Tài liệu & Office cho agent AI

GEO PDF Report Generator (pandoc pipeline)

Prerequisites

• pandoc — brew install pandoc
• Google Chrome — must be installed at /Applications/Google Chrome.app/

No Python dependencies. No ReportLab. No JSON data wrangling.

How It Works

1. Read GEO-AUDIT-REPORT.md in the current directory (created by /geo audit)

2. Extract cover metadata from the report (brand name, domain, GEO score, date, locations)

3. Run pandoc with separately provided reviewed CSS + HTML template to produce a self-contained GEO-REPORT.html

4. Run Chrome headless to print the HTML to GEO-REPORT.pdf

The pandoc template (references/ templates in this skill or user-provided HTML) injects:

• A full-bleed dark navy cover section with the GEO score badge
• Per-section cover metadata (date, business type, locations, platform)
• JavaScript that runs inside Chrome before printing to color-code score cells and severity-tag finding sections

Workflow

Step 1: Check for audit report

Look for GEO-AUDIT-REPORT.md in the current directory. If absent, tell the user to run /geo audit <url> first.

Step 2: Extract cover metadata from the report

Read the top of GEO-AUDIT-REPORT.md and extract:

| Field | Where to find it |

|---|---|

| brand_name | First H1 title (after "GEO Audit Report:") |

| domain | Second bold line (e.g. Domain: alexamediasolutions.com) |

| geo_score | Line matching ## Overall GEO Score: XX / 100 |

| score_label | Word after the score on that same line (e.g. "Poor", "Fair", "Good") |

| date | Audit Date: line |

| business_type | Business Type: line |

| locations | Locations: line |

| platform | CMS: line |

Step 3: Run pandoc

Save extracted metadata with a structured JSON writer to geo-metadata.json. Treat report fields as untrusted values, never executable code or shell fragments. The templates are not bundled: require existing reviewed absolute template/CSS paths in that JSON, or stop and report the missing prerequisite.

import json
import subprocess
from pathlib import Path

config = json.loads(Path("geo-metadata.json").read_text())
args = ["pandoc", "GEO-AUDIT-REPORT.md", "--to", "html5", "--standalone"]
for option, key in [("--template", "template_path"), ("--css", "css_path")]:
    path = Path(config[key])
    if not path.is_absolute() or not path.is_file():
        raise ValueError("An existing reviewed absolute resource path is required")
    args.extend([option, str(path)])
allowed = {"title", "brand_name", "domain", "geo_score", "score_label", "date", "business_type", "locations", "platform"}
for key, value in config.get("metadata", {}).items():
    if key not in allowed or not isinstance(value, str) or len(value) > 500:
        raise ValueError("Invalid report metadata")
    args.extend(["--metadata", key + "=" + value])
subprocess.run(args + ["-o", "GEO-REPORT.html"], check=True, shell=False)

Render with network access disabled and the browser sandbox enabled; do not embed remote resources or run untrusted template JavaScript.

Step 4: Run Chrome headless

The command below does not itself disable network access. Enforce that separately using the host’s supported sandbox/network policy, choose an isolated temporary browser profile and a new output destination, and inspect the generated HTML for remote resources or active content before rendering.

"/Applications/Google Chrome.app/Contents/MacOS/Google Chrome" \
  --headless=new \
  --disable-gpu \
  --print-to-pdf="$(pwd)/GEO-REPORT.pdf" \
  --print-to-pdf-no-header \
  --no-pdf-header-footer \
  --virtual-time-budget=5000 \
  "file://$(pwd)/GEO-REPORT.html"

Step 5: Report completion

Tell the user:

• GEO-REPORT.pdf was generated in the current directory
• File size
• Optionally: open GEO-REPORT.pdf to preview it

What the PDF Contains

• Cover page — Dark navy gradient, brand name, domain, GEO score badge (colored by score), audit date, business type, locations, CMS platform
• Score tables — Cells containing XX/100 are color-coded: ≥80 green, ≥65 blue, ≥50 amber, ≥35 orange, <35 red
• Finding sections — h3 headings containing "Critical / High / Medium / Low" get severity-colored left-border callout blocks (red / orange / yellow / green)
• Section page breaks — Major sections (High Priority, 90-Day Roadmap, Component Score Summary, Generated Schema) break to new pages automatically
• Code blocks — JSON schema templates render with dark theme monospace styling
• Page footer — Brand name · GEO Audit · date + page numbers (via CSS @page)

Customizing the Report

• Colors / typography — Edit references/ templates in this skill or user-provided HTML
• Cover layout — Edit references/ templates in this skill or user-provided HTML
• Score thresholds for color-coding — Edit the scoreColor() function in the template's <script> block
• Which sections get page breaks — Edit the breakBefore array in the template's <script> block

Troubleshooting

| Problem | Fix |

|---|---|

| pandoc: command not found | brew install pandoc |

| Chrome not found | Check path: /Applications/Google Chrome.app/Contents/MacOS/Google Chrome |

| PDF is blank / empty | Increase --virtual-time-budget to 8000 |

| Cover metadata missing | Check GEO-AUDIT-REPORT.md has the standard header format |

| Fonts not loading | PDF is rendered offline; system fonts are used as fallback — this is expected |

When to Use

• You need a Generative Engine Optimization task for a website: audit, citability, crawlers, schema, llms.txt, content, platform tuning, or client reporting.
• Run read-only analysis first; propose site changes before making any.

Limitations

• Audits are read-only analysis; never publish, deploy, or modify the target site without explicit approval.
• Scores and citation likelihoods are heuristics, not guarantees from AI search platforms.
• Docs-only import: upstream scripts, agents, hooks, and schema templates are not bundled.

Example

curl -s https://example.com/robots.txt
curl -s https://example.com/llms.txt

> Adapted from [zubair-trabzada/geo-seo-claude](https://github.com/zubair-trabzada/geo-seo-claude) (MIT); frontmatter, When to Use/Limitations, and safety boundaries added for upstream compliance. Docs-only import: upstream runtime helpers not bundled.