Building a Blog (Next.js + Sanity)
Overview
Universal workflow for adding a blog to a Next.js site backed by Sanity CMS. Targets a corporate blog: SEO-first, i18n-ready, performance-budgeted, accessibility-compliant.
SKILL.md is the orchestrator. The spec lives in the two reference files.
Files
blog-technical-requirements.md— §0 intake questionnaire (30+ questions) → §1 Project Profile → §2–§20 universal spec → checklist. Edit only §0 answers and §1 per project.blog-image-style-guide.md— universal template for AI-generated hero imagery via Gemini 3 Pro Image (Nano Banana Pro). Intake questions on top, aesthetic skeleton in the middle, three example slots at the bottom.
Workflow
Step 1 — Scan the host project (before asking anything)
Read what's already there. Pre-filling answers from detected values is much better than asking the user. Look for:
package.json— Next.js version,next-intl, Sanity packages, Tailwind,framer-motionnext.config.{ts,js,mjs}— existingimages.formats, i18n configsrc/messages/**,src/i18n/**,messages/**— current locale setsrc/app/[locale]/**orapp/[locale]/**— routing patterntailwind.config.*,globals.css, design-system files — tokens, fontsCLAUDE.md,AGENTS.md,GEMINI.md— project conventions and approved overrides- Existing
sanity-studio/orstudio/folder app/layout.{tsx,jsx}— fonts, analytics, motion configpublic/brand/**,public/logo*— publisher logositemap.ts,robots.ts— existing routes.env*files — names of already-set env vars (do not log secret values)
Record findings as a brief "Detected" list before the questionnaire.
Step 2 — Run the intake questionnaire
Open blog-technical-requirements.md and walk through §0.
- Claude Code: use
AskUserQuestion. Group related questions into single calls (max 4 per call). Pre-fill recommended answers from the scan. - Codex / CLIs without an interactive picker: list questions numerically in plain text, ask the user to answer in batches of 5–10. Show recommended answers from the scan.
- Headless / non-interactive: apply universal defaults to whatever the prompt did not explicitly cover; list every assumption.
Write the answers into §1 "Project Profile" (in a project-local copy under docs/blog/, not in the universal source).
Step 3 — Produce the high-level plan
Use the plan template at the end of §0. One page. Phases, locked-in scope, out-of-scope items, open decisions. Wait for explicit user approval. Do not start coding.
Step 4 — Implement against the spec
Follow §2–§20 in order. §19 (Pass/Fail Checklist) is the definition of done. The image style guide drives §20 if AI-generated hero images are in scope.
When NOT to use this skill
- Static / MDX-only blogs (no CMS) — different stack, different patterns
- High-velocity news publishers — sitemap chunking and editor tooling become primary concerns
- Documentation sites — use a docs framework instead
- Marketing landing pages that aren't really a blog
Source
Maintained at BuildShipGrowRepeat/nextjs-sanity-blog-skill. MIT licensed.
Building a Blog (Next.js + Sanity)
Overview
Universal workflow for adding a blog to a Next.js site backed by Sanity CMS. Targets a corporate blog: SEO-first, i18n-ready, performance-budgeted, accessibility-compliant.
SKILL.md is the orchestrator. The spec lives in the two reference files.
Files
blog-technical-requirements.md— §0 intake questionnaire (30+ questions) → §1 Project Profile → §2–§20 universal spec → checklist. Edit only §0 answers and §1 per project.blog-image-style-guide.md— universal template for AI-generated hero imagery via Gemini 3 Pro Image (Nano Banana Pro). Intake questions on top, aesthetic skeleton in the middle, three example slots at the bottom.
Workflow
Step 1 — Scan the host project (before asking anything)
Read what's already there. Pre-filling answers from detected values is much better than asking the user. Look for:
package.json— Next.js version,next-intl, Sanity packages, Tailwind,framer-motionnext.config.{ts,js,mjs}— existingimages.formats, i18n configsrc/messages/**,src/i18n/**,messages/**— current locale setsrc/app/[locale]/**orapp/[locale]/**— routing patterntailwind.config.*,globals.css, design-system files — tokens, fontsCLAUDE.md,AGENTS.md,GEMINI.md— project conventions and approved overrides- Existing
sanity-studio/orstudio/folder app/layout.{tsx,jsx}— fonts, analytics, motion configpublic/brand/**,public/logo*— publisher logositemap.ts,robots.ts— existing routes.env*files — names of already-set env vars (do not log secret values)
Record findings as a brief "Detected" list before the questionnaire.
Step 2 — Run the intake questionnaire
Open blog-technical-requirements.md and walk through §0.
- Claude Code: use
AskUserQuestion. Group related questions into single calls (max 4 per call). Pre-fill recommended answers from the scan. - Codex / CLIs without an interactive picker: list questions numerically in plain text, ask the user to answer in batches of 5–10. Show recommended answers from the scan.
- Headless / non-interactive: apply universal defaults to whatever the prompt did not explicitly cover; list every assumption.
Write the answers into §1 "Project Profile" (in a project-local copy under docs/blog/, not in the universal source).
Step 3 — Produce the high-level plan
Use the plan template at the end of §0. One page. Phases, locked-in scope, out-of-scope items, open decisions. Wait for explicit user approval. Do not start coding.
Step 4 — Implement against the spec
Follow §2–§20 in order. §19 (Pass/Fail Checklist) is the definition of done. The image style guide drives §20 if AI-generated hero images are in scope.
When NOT to use this skill
- Static / MDX-only blogs (no CMS) — different stack, different patterns
- High-velocity news publishers — sitemap chunking and editor tooling become primary concerns
- Documentation sites — use a docs framework instead
- Marketing landing pages that aren't really a blog
Source
Maintained at BuildShipGrowRepeat/nextjs-sanity-blog-skill. MIT licensed.
Scanned by
NVIDIA SkillSpector
Automated static analysis (no LLM) · v2.12.0 · 2026-11-19
Partial analysis: some checks could not run on this component.
4 observations in 2 categories
- Analysis evasion 2
- Excessive Agency 2
SkillSpector looks for patterns such as shell commands, credential access, network calls or unpinned dependencies. An observation is something worth reading in the files before you install, not a confirmed vulnerability. Components that run code locally, like hooks, MCP servers and security tooling, naturally show more of them. Open one to see the lines it points at.
Incomplete referenced artifact analysis ×2Analysis evasion · SKILL.md:20
| 18 | ## Files |
| 19 | |
| 20 | - `blog-technical-requirements.md` — §0 intake questionnaire (30+ questions) → §1 Project Profile → §2–§20 universal spec → checklist. Edit only §0 answers and §1 per project. |
| 21 | - `blog-image-style-guide.md` — universal template for AI-generated hero imagery via Gemini 3 Pro Image (Nano Banana Pro). Intake questions on top, aesthetic skeleton in the middle, three example slots at the bottom. |
| 22 |
Autonomous Decision Making ×2Excessive Agency · blog-technical-requirements.md:89
| 87 | |
| 88 | H1. **Analytics tool** — Matomo / Plausible / Fathom / GA4 / Posthog / Umami / none. Self-hosted vs. SaaS. |
| 89 | H2. **Consent model** — cookieless tool (no banner required) / consent-gated (load only after the visitor accepts cookies via the existing site-wide gate) / no consent gate (only legal in jurisdictions where the tool is genuinely cookieless |
| 90 | |
| 91 | #### I. Content authoring & workflow (3) |
Learn more about NVIDIA SkillSpector or how components are reviewed.
Agent compatibility
Checked by NVIDIA SkillEvaluator Harness Check · v1.0.1 · 2026-11-18
How each agent reads this skill once it is installed, from the agents' documented loading rules. No agent was run. A note means the agent loads the skill but changes part of its text, or drops it.
Claude Code
loadsCodex
loadsCopilot CLI
loadsOpenCode
loadsHermes
loadsOpenClaw
loads
Every agent loads this skill as written. No notes.
Learn more about NVIDIA SkillEvaluator.