content-pipeline-skill

Built by Artur Ferreira @ The GEO Lab · 𝕏 @TheGEO_Lab · LinkedIn · Reddit

Multi-agent content production pipeline for Claude Code where 7 specialist agents work together to research, write, edit, optimise for SEO/GEO, and validate against live analytics — with a master agent reviewing all outputs before human approval.

Who This Is For

• Content teams who want AI-assisted production with human quality gates
• Solo creators who need research → writing → editing → SEO in one workflow
• SEO practitioners who want content optimised for both traditional search and AI citation
• Anyone building on Claude Code who wants a reference architecture for multi-agent pipelines

What Makes This Different

Most “content AI” tools generate text and call it done. This skill orchestrates a full pipeline:

• ✅ 7 specialist agents — research, writer, editor, SEO/GEO optimizer, analytics, GEO quality gate, master QA gate
• ✅ GEO Quality Gate — 5-question per-section checklist ensuring every H2 is citation-ready, pronoun-resolved, single-topic, data-backed, and opens with a direct answer
• ✅ Agent handoff protocols — structured context passing between agents with token budgets
• ✅ Quality checkpoints — gates between every pipeline stage, not just at the end
• ✅ Analytics integration — pulls live GSC + GA4 data to inform content decisions
• ✅ GEO optimisation — not just SEO keywords, but AI extractability and citation readiness
• ✅ Human approval gate — pipeline halts before publishing, outputs to .claude/pipeline/
• ✅ Slash command — trigger with /content-pipeline inside Claude Code
• ✅ Production-tested — part of the The GEO Lab toolkit for AI-assisted content operations

Pipeline Flow

Research Agent → Writer Agent → Editor Agent → SEO/GEO Agent
                                                     ↓
                                              Analytics Agent
                                                     ↓
                                              Master Agent (GEO Quality Gate → GO / NEEDS REVISION)
                                                     ↓
                                              Human Approval → Publish

Install

# Clone
git clone https://github.com/arturseo-geo/content-pipeline-skill.git ~/.claude/skills/content-pipeline

# Or install all skills at once
git clone https://github.com/arturseo-geo/claude-code-skills.git
cp -r claude-code-skills/skills/content-pipeline ~/.claude/skills/

File Structure

content-pipeline-skill/
├── SKILL.md                  — Core skill instructions and pipeline orchestration
├── agents/
│   ├── master-agent.md       — Final QA gate: runs GEO Quality Gate, reviews all outputs, GO/NEEDS REVISION
│   ├── geo-quality-gate.md   — 5-question per-section GEO compliance checklist (run by Master Agent as Step 0)
│   ├── research-agent.md     — Web search, topic analysis, competitor content
│   ├── writer-agent.md       — Draft creation from research + analytics briefs
│   ├── editor-agent.md       — Quality, accuracy, brand voice, humanisation
│   ├── seo-geo-agent.md      — Keyword optimisation, schema, GEO readiness
│   └── analytics-agent.md    — GSC + GA4 + DataForSEO data pull
├── commands/
│   └── content-pipeline.md   — Slash command: /content-pipeline
├── references/
│   ├── agent-handoffs.md     — Agent-to-agent context passing and token budgets
│   ├── quality-gates.md      — Quality checkpoints between pipeline stages
│   └── analytics-integration.md — Data source setup and fallback behaviour
└── .github/                  — Issue templates and PR template

GEO Quality Gate (new in v1.1.0)

The Master Agent now runs a mandatory 5-question GEO quality gate on every H2 section before any other review checks:

1. Direct opener — Does the section open with a declarative answer, not context-dependent preamble?
2. Pronoun resolution — Are all pronouns replaced with explicit named entities?
3. Single topic — Does every paragraph serve one concept only?
4. Verifiable data — Is there at least one cited statistic or named study?
5. Citation-ready — Can the opening sentence stand alone as a quote with author attribution?

Any section failure triggers NEEDS REVISION with specific rewrite instructions for the Writer Agent.

Related Repos

• claude-code-skills — Full collection of skills
• seo-geo-skill — The SEO/GEO skill used by the pipeline’s optimizer agent
• mcp-wordpress-setup — WordPress MCP server for publishing pipeline output

Contributing

See CONTRIBUTING.md for guidelines. PRs welcome.

---

Built and maintained by Artur Ferreira · Part of the The GEO Lab toolkit · MIT License