Skip to content

Sanity Toolkit Skills Evaluation

This report evaluates the agent skills provided in the sanity-io/agent-toolkit repository for their potential use within the Zero-Shot Agency project.

1. Utility

The repository provides six major skills primarily aimed at structuring, migrating, and presenting content within headless CMS environments (specifically Sanity).

The genuinely useful skills include: - seo-aeo-best-practices: Direct utility for Generative Engine Optimization (GEO). Provides guidelines on AI-answer engine readiness, EEAT principles, and crawler management. - content-modeling-best-practices: Highly useful for building out robust schema designs and separating content from presentation. As we are curating content for LLM ingestion, strict structural content modeling is crucial. - content-experimentation-best-practices: Useful for establishing A/B testing frameworks and understanding statistical significance for content variants, which can be applied to test GEO tactics across different LLM targets.

The remaining skills (portable-text-conversion, portable-text-serialization, sanity-best-practices) are highly specific to the Sanity ecosystem and Portable Text specification, offering little utility to our current MkDocs + Markdown stack unless a migration is planned.

2. Evidence

The sanity-io toolkit skills are rigorously documented and heavily backed by proven workflows rather than guesswork: - content-experimentation-best-practices references concrete statistical foundations (p-values, Bayesian methods) and common pitfalls. - portable-text-conversion & serialization are supported by explicit Sanity libraries (@portabletext/markdown, @portabletext/block-tools) with concrete component mapping patterns across various modern frameworks (React, Astro, Svelte). - seo-aeo-best-practices is deeply grounded in Google's established EEAT framework and JSON-LD schema guidelines, and provides structured checklists for technical SEO.

3. Application to Zero-Shot Agency

For the current Zero-Shot Agency project, we should directly apply the principles from the following skills:

  1. AEO & EEAT Implementation: We must integrate the guidelines from seo-aeo-best-practices into our geo-tactics. Ensuring our content adheres to AI Overview constraints, structured data (JSON-LD), and crawler management is vital to being cited by Perplexity, ChatGPT, Claude, and Gemini.
  2. Content Architecture: We should adopt principles from content-modeling-best-practices to refine our [[SCHEMA]] and llms.txt definitions. Treating "content as data, not pages" aligns perfectly with creating LLM-native assets and maintaining a "single source of truth".
  3. Experimentation Framework: We can adapt the hypothesis framework from content-experimentation-best-practices to measure the citation success rates of various structural formats (e.g., Markdown vs. Semantic HTML) within our geo-tracker tool.

Unless we intend to migrate Zero-Shot Agency from MkDocs to Sanity, we should skip the implementation of sanity-best-practices and Portable Text skills.