exec-meta-gen — Bulk SEO/AEO/GEO Meta Generator

Bulk-generate optimised title tags, meta descriptions, and OG tags for a list of pages from a CSV. Powered by Claude Haiku. Built as an Abacus.ai ChatLLM agent tool and a standalone Python CLI.

Companion to exec-meta-eval — generates the meta, then scores it so you know quality before publishing.

---

What It Does

For each row in your input CSV (url, keyword):

1. Fetches the live page to extract existing content for context
2. Generates optimised meta via Claude Haiku:
   • Title tag (50–60 chars, keyword near front)
   • Meta description (145–160 chars, keyword + value hook + CTA)
   • OG title (benefit-led, ≤70 chars)
   • OG description (outcome-focused, ≤200 chars)
3. Scores the generated meta with the exec-meta-eval 9-dimension rubric (0–100)
4. Returns a CSV ready to paste into your CMS or bulk-upload sheet

---

Input Format

Minimum required columns:

url,keyword
https://example.com/solutions,AI-led business transformation
https://example.com/insurance,insurance claims automation

Optional columns for better output:

url,keyword,page_type,audience
https://example.com/solutions,AI transformation,Landing Page,CMO
https://example.com/case-study,claims automation ROI,Case Study,CFO

Column	Values
page_type	Home, Landing Page, Product/Solution, Case Study, Blog
audience	CMO, CTO, CFO, VP Operations, etc. (defaults to "B2B senior decision-maker")

---

Output CSV

Column	Description
url	Input URL
keyword	Input keyword
title	Generated title tag
title_len	Character count
meta_desc	Generated meta description
meta_desc_len	Character count
og_title	Generated OG title
og_desc	Generated OG description
reasoning	Why this keyword placement and CXO hook was chosen
eval_score	exec-meta-eval score (0–100)
eval_grade	A / B / C / D / F
top_fix	Single highest-impact improvement if score < 70
error	Error message if fetch or generation failed

---

Usage

Abacus.ai ChatLLM Agent

1. Upload agent_tool.py as a code tool in your Abacus.ai ChatLLM agent
2. Set ANTHROPIC_API_KEY in the agent's environment variables
3. Copy the system prompt from agent_instructions.md into the agent instructions
4. Connect generate_meta_bulk as the callable function

See agent_instructions.md for the full setup guide and example interaction.

CLI (local / any Python environment)

pip install httpx
export ANTHROPIC_API_KEY=sk-ant-...

python agent_tool.py input.csv output.csv

Python API

from agent_tool import generate_meta_bulk

csv_text = """url,keyword,page_type,audience
https://example.com/solutions,AI transformation,Landing Page,CMO
https://example.com/insurance,claims automation,Product,CFO"""

result_csv = generate_meta_bulk(
    csv_input=csv_text,
    api_key="sk-ant-...",
    fetch_pages=True,
    include_eval=True,
)
print(result_csv)

---

Cost Estimate

Each row makes 2 Claude Haiku calls (generate + eval) + 1 HTTP fetch. Haiku pricing at time of writing: ~$0.25 / 1M input tokens, ~$1.25 / 1M output tokens.

Rows	Approx. cost	Approx. time
10	~$0.01	~30s
100	~$0.10	~3–4 min
500	~$0.50	~15–20 min

(With fetch_pages=False and include_eval=False, cost and time drop by ~60%.)

---

Files

exec-meta-gen/
├── agent_tool.py           # Core Python tool + CLI entry point
├── agent_instructions.md   # Abacus.ai ChatLLM system prompt + setup guide
├── requirements.txt
└── README.md

---

Related Skills

• exec-meta-eval — evaluate existing meta
• content-eval — evaluate long-form content

---

Built by RYVR Immersive.