Tech
Workflow Process
- 1
SEO Audit & Content Discovery
Analyzes existing content and searches for new keyword opportunities. Based on the results, the pipeline either updates existing pages or creates entirely new content.
- Content Audit Parses every content file and cross-references live Ahrefs ranking data to flag weak titles, thin pages, and missing keywords
- Gap Discovery 3 parallel Ahrefs MCP keyword queries returning 150 results, filtered through multi-layer deduplication
- Decision Existing pages with SEO issues get rewritten, new keyword gaps trigger full content creation
- 2
Competitor Research
Playwright MCP navigates top-ranking pages and extracts content structure through browser snapshots.
- DOM Extraction Captures headings, section structure, and content patterns from competitor pages
- Pattern Signal Extracted structure shapes the generated content's depth and format
- 3
Content Generation
Generates new content or rewrites existing pages based on audit findings and competitor research patterns.
- Structured Output Multiple content variants designed upfront with category-aware image classification
- SEO-Driven Incorporates target keywords, heading structure, and competitor insights into every piece
- 4
File Scaffolding
Creates structured files per content piece with a slug-based safety check before writing to disk.
- Output index.md with TOML frontmatter, content JS with interactive data, metadata JS with display settings
- Safety Mandatory slug collision check prevents overwriting existing content
- 5
Image Generation
Feature images via Freepik's Mystic API with async polling and automatic compression.
- Flow POST to generate → poll task_id → download → compress to 100-500 KB
- Fallback Flags manual reminder if API is unavailable
- 6
Parallel Translation
20 concurrent sub-agents launched in a single message, one per target language.
- Scope Each agent translates 2 files; metadata and images inherited by Hugo
- Quote Safety Straight-quote enforcement in JS files to prevent Hugo minifier failures
- 7
Verification
Multi-directory validation across all 20 language directories.
- Checks Confirms file count and parses JS syntax in every language directory
- Recovery Failed languages re-run individually without restarting the pipeline
From keyword search to validated content in 20 languages. A single skill invocation, zero human intervention, 60+ files generated per run.