rankm8-content-pipeline

name	rankm8-content-pipeline
description	Use this skill when a user wants to create SEO analysis plans, outlines, full content, meta information, or AI images with RankM8 MCP generation tools.

RankM8 Content Generation

Purpose

Create SEO content from RankM8 project data in a controlled sequence: analysis plan, outline, content, meta information, and optional image generation. Use real pages, keywords, SERPs, and competitor context whenever possible.

Required MCP Tools

• generate_analyse_plan(project_id, prompt?, model?, page_ids?, keyword_ids?, settings?)

• generate_outline(project_id, prompt?, model?, page_ids?, keyword_ids?, plan_id?, settings?)

• generate_content(project_id, prompt?, model?, page_ids?, keyword_ids?, plan_id?, outline_id?, settings?)

• generate_full_content(project_id, prompt?, model?, page_ids?, keyword_ids?, settings?)

• generate_meta_info(project_id, content_id?, prompt?, model?, page_ids?, keyword_ids?)

• generate_image(prompt, project_id, reference_image_base64?, reference_image_urls?, reference_image_ids?, model?)

• get_generation_status(entity_type, entity_id)

• list_projects(limit?, offset?)

• list_pages(project_id, search?, limit?, offset?)

• search_keywords(project_id, query?, limit?)

• get_keyword_details(keyword_id)

• keyword_research(keyword, type, project_id?, language?, location_code?, include_serp_info?, limit?)

• analyze_keywords(project_id, keyword_ids)

• list_models(category?, capability?)

Step-by-Step Workflow

1. Identify and confirm project_id. Never auto-select a project for generation.
2. Ask the user for the topic, target page, target keywords, desired length, and tone if they were not provided.
3. Use list_pages(project_id, search?, limit?, offset?) to find target pages and confirm page_ids.
4. Use search_keywords(project_id, query?, limit?) to find target keywords and confirm keyword_ids.
5. For each target keyword, check keyword_fit. If it is null, ask whether to improve data quality before writing.
6. If the user approves, run analyze_keywords(project_id, keyword_ids=[...]) and poll get_keyword_details(keyword_id) every 30 seconds until keyword_fit is no longer null or the user asks to continue without full analysis.
7. Optionally call list_models(category="chat") when the user asks for a model or quality/cost tradeoff.
8. Confirm credit usage for generate_analyse_plan.
9. Call generate_analyse_plan(project_id, prompt?, model?, page_ids?, keyword_ids?, settings?).
10. Poll get_generation_status(entity_type="plan", entity_id=writing_plan_id) every 30 seconds until completed or failed.
11. Review the completed plan with the user before creating an outline.
12. Confirm credit usage for generate_outline.
13. Call generate_outline(project_id, prompt?, model?, page_ids?, keyword_ids?, plan_id=writing_plan_id, settings?).
14. Poll get_generation_status(entity_type="outline", entity_id=writing_outline_id) every 30 seconds until completed or failed.
15. Review the outline with the user and ask for approval before full content.
16. Confirm credit usage for generate_content.
17. Call generate_content(project_id, prompt?, model?, page_ids?, keyword_ids?, plan_id=writing_plan_id, outline_id=writing_outline_id, settings?).
18. Poll get_generation_status(entity_type="content", entity_id=writing_content_id) every 30 seconds until completed or failed.
19. Confirm credit usage for generate_meta_info.
20. Call generate_meta_info(project_id, content_id=writing_content_id, prompt?, model?, page_ids?, keyword_ids?). This is synchronous; no polling is needed.
21. If an image is requested, ask for image direction, references, and final project confirmation.
22. Confirm credit usage for generate_image.
23. Call generate_image(prompt, project_id, reference_image_base64?, reference_image_urls?, reference_image_ids?, model?).

Polling Instructions

For generation jobs, always use get_generation_status(entity_type, entity_id). Use entity_type="plan" for writing_plan_id, entity_type="outline" for writing_outline_id, and entity_type="content" for writing_content_id.

Valid generation statuses are pending, processing, completed, and failed. Poll every 30 seconds until terminal. If a response contains poll_again_in_seconds, follow it. If a response contains warning or times out to failed with failure_reason="timeout", stop and ask before starting a new job.

generate_full_content is an alternative single-call pipeline. It returns writing_plan_id, writing_outline_id, and writing_content_id; poll the content entity for the whole pipeline and optionally poll plan/outline for intermediate artifacts.

User Confirmation Points

Always confirm before:

• Calling any credit-consuming generation tool.
• Using a selected project for generated assets or content.
• Running analyze_keywords to improve keyword quality.
• Continuing from plan to outline and from outline to content.
• Using reference images with generate_image.

No delete tools are part of this workflow. If the user asks to remove generated content from RankM8, explain that this playbook does not expose a delete step.

Error Handling

• insufficient_credits: Stop and ask the user to add credits or choose a cheaper step.
• invalid_model: Call list_models(category="chat") for text or list_models(category="image-generation") for images.
• project_not_found: Re-select the project or reconnect OAuth.
• plan_not_found, outline_not_found, or content_not_found: Re-poll or use IDs from the current project only.
• invalid_settings: Correct settings values like content_length, custom_word_count, writing_style, topical_terms_optimization, analysis_depth, or section_depth.
• missing_prompt: Provide a prompt for generate_image.
• generation_failed or reference_image_failed: Show the error details and retry only after the user confirms changed inputs.
• not_found: Treat as access denied or stale ID; re-select the artifact.