Agent recipes

The acceptance recipe: a fresh Claude Code session connects and reads visibility in one prompt.

MCP tools: get_data_freshness · get_llm_visibility · get_visibility_score

Paste this into your agent:

Summarize my AI visibility: how fresh is the data, my composite Visibility Score and its pillars, and where my brand is (not) cited across ChatGPT/Perplexity/Gemini/Claude/Google AI. Flag any model where I'm absent.

See which URLs/sources AI models cite when they mention your brand, and which competitors appear instead.

MCP tools: get_llm_responses · get_competitor_intel

Paste this into your agent:

Pull recent AI model responses for my domain. List the sources each model cites, where my brand is positioned, and which competitors show up instead of me. Rank competitors by frequency.

Turn pooled opportunities + briefs into a 30-day plan.

MCP tools: get_content_calendar · list_content_opportunities · list_content_briefs · get_keyword_pool

Paste this into your agent:

Draft a 30-day content plan. Pull my current calendar, queued opportunities, and ready briefs; prioritize by opportunity score and keyword difficulty; avoid cannibalizing existing articles. Output as a week-by-week table with keyword, intent, and one-line brief per article.

Striking-distance keywords, low-CTR pages, and declining pages.

MCP tools: get_gsc_opportunities · list_low_ctr_queries · list_declining_pages

Paste this into your agent:

Find my top 10 quick-win SEO fixes: striking-distance keywords (positions 5–15), queries with strong impressions but low CTR, and pages declining vs last period. For each, suggest one concrete action (refresh, title tweak, internal link).

Period-over-period GSC delta.

MCP tools: compare_gsc_periods · get_site_performance_summary

Paste this into your agent:

Compare my GSC performance this month vs last month: clicks, impressions, CTR, average position. Highlight the biggest movers (both directions) and call out any metric that regressed.

Write-scoped: publish or reschedule without leaving the agent.

MCP tools: get_articles · publish_article · reschedule_article · set_article_status

Paste this into your agent:

List my generated-but-unpublished articles. For the one titled "<match>", publish it to its configured destination. If it's scheduled for the wrong day, reschedule it to YYYY-MM-DD instead.

Find refresh candidates + cannibalization risk before rewriting.

MCP tools: get_refresh_candidates · get_cannibalization · get_article_details

Paste this into your agent:

List articles queued for refresh with their decay reason. For the top 3, pull full content and the cannibalization signals, then propose a refresh outline that protects the existing keyword cluster rather than cannibalizing it.

Foundational audit: schema, entity, platform sub-scores + findings.

MCP tools: get_geo_health · get_seo_health · get_serp_features

Paste this into your agent:

Run a GEO/AI-readiness audit: schema/entity/platform sub-scores, top findings, and which pooled keywords trigger AI Overviews / featured snippets / PAA. Prioritize the findings the coding agent could fix as PRs.

For agents that crawl docs directly, Rankealo publishes these at the site root:

• /llms-full.txt — the full consolidated doc corpus for LLMs.
• /llms.txt — the lightweight index.
• /agent-sitemap.xml — an agent-oriented sitemap.

Point an agent at /llms-full.txt to give it the whole doc set in one fetch.

For Claude Code users, Rankealo ships an official Skill that wraps these recipes. Drop SKILL.md into your project's .claude/skills/rankealo/ directory (or install via the MCP init flow) and the three core jobs — check my AI visibility, list my top cited sources, draft next month's content plan — become first-class slash-invocable workflows. The skill source lives at mcp-server/skills/rankealo/SKILL.md in the Rankealo repo.