Learning path map
Detailed content
💻 Onboarding + Audit Report
The pair of skills that opens a consulting project: generates the complete client onboarding package and turns your audit notes into a professional AI Readiness Report.
A standardized package—scope agreement, questionnaire, access checklist, and timeline—that every new project needs before the first line of work.
It's the consultant's most repetitive deliverable. Automating it frees up hours per client and creates a professional first impression.
Scope, stakeholders, access, expectations, timeline—the five pillars of any kickoff.
The artifacts the skill produces: welcome doc, discovery questionnaire, list of credentials to collect, and week 1 plan.
Understanding each artifact lets you adapt it to your niche without reinventing the structure.
Template, placeholder, client variable, generated document.
The second skill takes raw audit notes and builds an AI readiness report with a score, findings, and a roadmap.
It's the deliverable that justifies the value of the consulting work — it turns observation into actionable recommendations.
Readiness score, finding, impact, effort, prioritized roadmap.
The dimensions assessed: data, processes, team, infrastructure, and use cases that can be prioritized.
Gives you a reusable framework for any company, instead of a standalone opinion.
Data maturity, automation, leverage, quick win.
Onboarding kicks off the project and gathers context; the Audit Report completes the diagnosis. The output of one becomes the input for the other.
Chained skills make up a complete consulting product, not two standalone utilities.
Chaining, handoff, shared context, productization.
Write a SKILL.md that generates the package and report with your branding, your fields, and your tone.
The generic skill is the starting point; what sets it apart is its specialization.
Frontmatter, triggers, versioned template, references/.
📊 SEO + AI Search Auditor
A skill that audits any site for traditional SEO AND visibility in AI search (AEO) in one pass — robots.txt, llms.txt, schema, Core Web Vitals, and a prioritized action plan.
SEO targets Google; AEO (Answer Engine Optimization) targets ChatGPT, Perplexity, and Gemini. The skill does both at the same time.
Most tools ignore one of the sides. Covering both is the market differentiator.
SEO, AEO, answer engine, AI citation.
Checking which AI crawlers (GPTBot, ClaudeBot, PerplexityBot) the site is unknowingly blocking.
One wrong rule in robots.txt and you never appear in AI answers — it’s the costliest and most common mistake.
User-agent, Disallow, AI crawler, high priority.
A machine-readable file that tells AIs where your authoritative content lives. The skill generates llms.txt e llms-full.txt ready.
It's the highest-impact AEO action that most sites still haven't taken.
llms.txt, llms-full.txt, authority, deploy.
JSON-LD (FAQPage, Organization, Article) and the answer-block method—question in the H2, direct answer right below.
Structure is what enables AI to confidently extract and cite your content.
JSON-LD, FAQPage, answer block, inverted pyramid.
The sequence the skill runs: finds core files, checks bots, detects the schema, and generates a prioritized report.
Understanding the workflow lets you trust the result and extend it (parallel competitor analysis).
web_fetch, subagents, quick win, prioritized plan.
The structure of the actual skill: SKILL.md with the playbook, audit.sh, an llms.txt generator, and three reference guides.
Shows progressive disclosure in practice: SKILL.md stays concise and delegates to the references.
scripts/, references/, decision tree, progressive disclosure.
🚀 RAG Architect
The skill that looks at your data, chooses among four RAG architectures, and explains why—before you write a line. It delivers a plan and runnable starter code, not stubs.
The core idea: treating all data the same (“chunk, embed, search”) fails when the data gets complex. The architecture depends on the data.
Choosing the wrong architecture is the #1 cause of RAG retrieving the wrong information.
RAG, retrieval, chunk, embedding, data-driven architecture.
It’s for anyone connecting data to an LLM—you don’t even need to know what RAG means. You describe the data; the skill decides.
Positions the skill as a consulting tool: the client talks business, it talks architecture.
Data signals, cardinality, scale, query pattern.
Naive (simple data), Advanced (hybrid search + exact codes), Modular/Agentic (multi-source with a router), and Graph (relationships between entities).
Knowing all four and their triggers is the heart of the architecture decision.
Hybrid search, query router, separate indexes, knowledge graph.
Chunk size justified by the architecture, embedding model, vector database choice, and retrieval method.
Magic numbers (chunk, top-k) are only useful if you know why they’re there and how to adjust them.
Chunk size, overlap, top-k, reranking, pgvector.
The skill generates ingest.ts, query.ts, config.ts with complete SQL migrations and a README that serves as a setup guide.
Code that runs on the first try is what separates a useful skill from a boilerplate generator.
Scaffold, migrations, no stub, stack awareness.
Write a SKILL.md that guides the 5 phases — understand the data, recommend, plan, scaffold, build — with the four architectures as knowledge.
It's the most advanced example in the learning path: a skill that makes engineering judgments, not just generates text.
Phased workflow, justified recommendation, companion build.