A brief for articles with no keyword (news, opinion, tutorials, build stories) built from your notes, an inspiration link, your experience and your stances.
---
name: "thought-leadership-brief"
description: "Write a brief for an article with no target keyword (news, opinion, a tutorial, a build story) from your notes, an inspiration URL, your experience and your stances. Same shape as the SEO brief, so the same research and writer skills run after it."
---
# Thought leadership brief (your site)
> **Shared from the growthvibes.ai content pipeline.** [How the whole pipeline works](https://growthvibes.ai/resources/how-i-built-an-seo-content-automation/). This is the skill we run, with our IDs, keys and private details taken out. Anything in angle brackets (`<like this>`) is yours to fill in. It is built to run unattended: it never asks questions, it decides and writes its assumptions down.
## What you need
- **Notion MCP** with a Briefs database, an Experience bank (your real stories) and a Positions log (your stances). Plain Markdown files work too if you point the skill at them.
- **Firecrawl MCP** to read the inspiration URL and run one search.
- **Optional:** the `content-library` skill (search your own saved reading) and your CMS connector for internal links.
- An idea: a title, your notes and optionally an inspiration URL. In an automation they come in as `work/<slug>/idea.json`; by hand, just paste them in.
## Fill in before the first run
- Database ids and your core pages, wherever placeholders appear.
Built for an automation (a runner picks a calendar row whose Track is Thought leadership), and works just as well by hand. Nobody is watching, so **never ask questions**. Decide, log it under Assumptions, keep going.
These articles do not chase a keyword and do not mirror the search results. They are built from an idea, the **inspiration URL** (the video or article the author wants something similar to), the author's own experience and positions, and the content library. The research skill and the writer run after this skill exactly as for SEO articles, so the brief must have the same shape.
**This skill writes the brief only.** The runner has already decided the track. Never second-guess it.
## Config
Same as `seo-content-brief`: site `your site`, articles at `/resources/`, author the author, Sanity project `<your Sanity project id>` dataset `production`, Notion Briefs data source `<your Notion data source>`, Content Calendar `<your Notion data source>`. Raw files go in the git-ignored `work/<slug>/`. Placeholder-page rules are the same as in `seo-content-brief`: never link to or use placeholder posts.
Do not use Ahrefs, AirOps, DataForSEO or Search Console. Do not scrape the top 10 results. Do not collect People Also Ask.
## Inputs
`work/<slug>/idea.json`, written by the runner: `title`, `notes`, `topic`, `category`, `idea_url`, `links_up_to` (the core page to link to) and `links_sideways` (same-category map pages; only link the ones whose status is `published`, they are live). If the file is missing, use the calendar row.
## Build-in-public articles ("how I built X")
If the idea's Notes or page body names a build log (a path like `work/build-logs/<name>.md`), the article is about something the author and Claude actually built, and the build log is the **primary first-hand source**. Then:
- Read the whole build log first. It lists the timeline, the stack, the steps, the mistakes, the numbers that are safe to use, a **do-not-disclose list** and what is UNKNOWN. Respect all of it.
- Set `build_in_public: true` in Brief data. This turns off invented first-person stories for the article: every "I did X" claim must trace to the build log, the repo (git history, docs, skills, scripts) or the experience bank. If a detail is UNKNOWN in the log, leave it out or write it as a plain gap ("I don't have the numbers on that"); never fill it.
- Treat every tool named in the build log as used by the author (`tested_by_george: true` in the entities), but only for what the log says was done with it.
- Check any claim in the title or notes against the log (for example "in one day" against the commit timestamps) and, if the log says otherwise, put the accurate version in the hook and log the change under Assumptions.
- Structure follows the build, in order, with the reasoning at each decision, what went wrong and what he would do differently. Each section ends with something the reader can copy: a rule, a prompt, a command, a template, a diagram. Include the do-not-disclose list as a check in the brief's Assumptions section.
- Length: these are deep, long articles. Cover every part of the process the build log describes; the notes may set a target (for example 4,000+ words). Go longer where the build needs it.
- Visuals: plan diagrams and timelines from the log's "visuals" list, and screenshots of public pages only.
## Steps
0. **Work out what the idea is made of.** An idea can be any mix of: a title only, the author's written notes (the calendar `Notes` property and the page body in `idea.json`: `notes`, `body`), and an `idea_url`. **the author's written notes are the strongest input**: they are his direction on what to cover, his angle and sometimes his thesis and outline. Keep his points, his order and his wording of the thesis, and build the rest around them. An inspiration URL, when there is one, adds material and structure but never overrides his notes. With only a title, derive the angle from the title plus his experience bank, positions and the library, and log under Assumptions that the angle was inferred.
1. **Understand the inspiration (only if there is an `idea_url`; otherwise skip to step 2).** Read it in full: Firecrawl scrape (markdown). For a YouTube video, the scrape often returns only the description: then search the library for it (`node <content-library skill>/scripts/search.mjs "<video title>" --limit 5`), because the author may have ingested the transcript. Extract: what it claims, its structure (the order of sections or steps), the examples and tools it names, and what it leaves out. If you can only get the title and description, say so under Assumptions and set status `low_confidence`. Never invent what a video says.
2. **Own material first (0 credits).**
- Experience bank (`<your Notion data source>`): rows whose Topics or tool match. These are the first-person stories and tool notes the writer will use. List each by URL and the section it fits.
- Positions log (`<your Notion data source>`): Active positions on this topic. The thesis should be consistent with them.
- Sanity: the author's published, non-placeholder posts and case studies on the topic.
- Content library: `node <content-library skill>/scripts/search.mjs "<question>" --limit 10` for the idea and for each outline section.
3. **One light look at what exists.** A single Firecrawl search on the idea (limit 5, no scraping, 2 credits) to see what is already written. It is for differentiation only: note the common angle and the one thing nobody says. It does not set the outline and nothing is mirrored. Skip it if the inspiration URL already shows the gap clearly.
4. **Choose the article type** (same values as the SEO brief: `explainer | tutorial | listicle | tool_review | comparison | service_roundup`; use `explainer` for news and opinion) and the **intent**: `Utility` (the reader will do something) or `Foundational` (the reader will understand something).
5. **Write the thesis**: one sentence, the position the article takes. It must be something the author can stand behind from his experience bank and positions. No thesis without an anchor in his material or a sourced fact.
6. **Build the outline.** Start from the author's notes if there are any, otherwise from the inspiration's structure, then go beyond it: add the sections it skips, the tradeoffs, when the advice does not apply, and the author's own experience where the bank has it. Cover every aspect of the idea; the article should be the most complete and honest treatment of it. Every H2 gets a one-line reason and the experience, library or position material that goes in it. Headings follow the hook and read like advice or questions, not keywords.
7. **Gaps for the research skill (2 to 5).** Things the inspiration and the existing articles do not settle: unverified claims, missing numbers, "does this still work today", costs, limits. Each gets 1 to 3 specific research questions. Items from the last 30 days belong to the research skill's recent step; name the repos, changelogs or pages it should check.
7b. **Visual plan.** Thought leadership pieces should be easy to follow, so plan the visuals in the brief, section by section (`visual_plan` in Brief data: `[{heading, shows, source, kind}]`). For each section ask what the reader must see to understand or do it (a setting, a dashboard screen, a code snippet's result, a chart, a flow). `source` is where a real image can come from, **public pages only**: the vendor's docs page that shows the screen (for example the docs page on where to find an API key), the product or changelog page, the GitHub README, a study, a post. Name the actual URL when you can find it (a scrape or one search; credits count toward the research cap). `kind` is `screenshot` (public page, crop and annotate), `evidence` (a passage or chart backing a claim), `library` or `own` (a branded flow, timeline, schematic or comparison we build). Logged-in screens (dashboards, Studio, Notion, Cloudflare) can never be captured: use the docs page that shows them, a schematic of our own, or, for build-in-public pieces, the screenshots the author put in the log's asset folder (`work/build-logs/assets/<log-name>/`, if it exists; look at each one and skip any that shows a key, ID or personal address). Make the article a `tutorial` type when its core is steps, so the research captures one image per key step.
8. **Subjects** (named tools or companies) with their home, pricing and docs URLs, and a `research_plan` (perspectives that matter, recency target of 30 days, visuals) in the same shape as the SEO brief.
9. **Title, meta, slug.** H1 is the hook, plain and specific, in the author's voice, not stuffed with a keyword. SEO title 50 to 60 characters and meta description 140 to 155 characters, both counted in code. Slug: lowercase, hyphens, short, from the topic. The "primary keyword" field holds the topic phrase (for example `ollama claude code`), and `secondary_keywords` stays empty or holds up to 3 natural phrases a reader might search. These are labels, not targets.
10. **Internal links.** Both core pages always, AI Automation & Agents and SEO & AEO (the one from `links_up_to` early, in the middle and in the closing; the other at least once), and the homepage, so every article has at least 4 internal links. After those, only relevant content: the live pages from `links_sideways`, plus any live, non-placeholder article or case study in Sanity (any category) that is directly about something a section covers (a worked example, a deeper guide on a subtopic, the same tool or system), each with the section it goes in, plus a service page where it fits. Sharing a category is not enough and there is no minimum: if you can't name the sentence a link belongs in, leave it out. Log dropped links and the ones added outside `links_sideways`.
11. **Length.** There is no minimum or maximum. Set `target` to your honest estimate of what covering the outline fully takes (about 250 to 350 words per H2), and say the writer should go longer where the topic needs it.
## Output (Notion only)
Same as `seo-content-brief` Phase 7:
- A brief page in the Briefs database (icon 📋) with Name = the working title, Status `Brief ready` (`Blocked` only if there is neither an idea title nor a URL), Confidence, Primary keyword (the topic phrase), Post type `New`, Cluster (the category), Cluster role `Spoke`, Intent, Target URL `https://your site/resources/<slug>/`, SEO title, Meta description, Target words. If a brief for this row already exists, update it instead of creating a second one.
- Link the brief in the calendar row's `Brief` property and set the row's Status to `Brief`. Never change the Publish date, `Idea` or `Live URL`.
- A child page "Brief data" with the JSON, in chunks of 2,000 characters labeled "Part X of Y".
Brief page sections (Notion markdown, real newlines): Status and Track, TL;DR, Thesis, Inspiration (what it covers, what it misses, URL), Core Pain Points, Why This Matters, Suggested Structure, SEO Title Meta Slug, Content Direction, Differentiation, Experience and Positions to use, FAQ (4 to 6 questions a reader would really ask, from the inspiration, the library and the research questions, each with a one-line direction), On-page and Schema, Internal Links, Citable Sources, Assumptions and Gaps.
Brief data JSON uses the same keys as `seo-content-brief` so the research skill reads it unchanged, with these differences: `track: "thought_leadership"`, `thesis`, `inspiration: {url, covers, misses}`, `visual_plan: [{heading, shows, source, kind}]`, `experience_to_use: [{url, section}]`, `positions_to_use`, `serp` set to `{"source": "none", "results": [], "paa": [], "related_searches": [], "ai_overview_cites": []}`, `competitors: []`, `competitor_keywords` empty, `coverage_map` built from the inspiration's sections (tier `must` for what the inspiration covers, `gap` for what you add), and `word_count: {"floor": null, "target": N, "target_reason": "outline estimate"}`.
## Rules
1. Never ask. Decide, log, continue.
2. Never invent what the inspiration says, or any number, quote or experience. Use only what was read, found in the author's material, or sourced.
3. The thesis and every experience claim trace to the experience bank, the positions log or published Sanity content.
4. No SERP scraping and no keyword targeting. The inspiration plus the author's material plus original research is the source.
5. Same Notion shape as the SEO brief, so research and writer run unchanged. One brief and one Brief data page per row, never duplicates.
6. Plain, practical language. No em dashes. No filler.
Not every article should chase a keyword. This brief is for the ones that shouldn’t: a new release, a strong opinion, a tutorial of something you built. It comes from my content automation and saves in the same shape as the SEO brief, so the same research and writer skills run after it.
What it does
Treats your written notes as the strongest input: your points, your order, your thesis
Reads the inspiration link (an article or a video) for structure and what it misses
Pulls matching stories from your experience bank and stances from your positions log
Runs one light search to see what is already written, for differentiation only
Plans the outline, the gaps to research and a visual plan for each section
Has a build-in-public mode: for “how I built it” pieces it turns off invented stories and uses your build log as the source
What you need
Notion MCP with a Briefs database, an Experience bank and a Positions log (or Markdown files)
Collect the idea. Title, notes, an optional inspiration URL.
Read your material first. Experience bank, positions, your published posts, the content library.
Look around once. One search to see the common angle and the thing nobody says.
Write the thesis. One sentence you can stand behind, anchored in your material.
Plan. Outline, gaps with research questions, a visual plan, links, a hook title.
Save. Same brief and JSON shape as the SEO brief.
Try asking
“Brief an article on Claude Code’s new release from these notes.”
“Turn this video into a brief for a how-I-do-it tutorial, my notes below.”
Install it
Download the ZIP above.
Claude Code: unzip it into .claude/skills/ in your project (or ~/.claude/skills/ for all projects). Keep the folder name; the scripts folder goes with it.
Add the keys from “What you need” to a .env file in your project, and connect the MCP servers it uses.
Claude.ai: upload the ZIP under Settings, Capabilities, Skills. The instructions work there too; the scripts need Claude Code.
Download the ZIP and unzip it into .claude/skills/ inside your project (or ~/.claude/skills/ for every project). Keep the folder name. The scripts folder inside it travels with the skill.
Can I use it in Claude.ai?
You can upload the ZIP under Settings, Capabilities, Skills, but the bundled scripts need a terminal, so Claude Code is the better home. The instructions still work in Claude.ai with the same connectors.
Is anything private left in it?
No. Our IDs, keys, database links and private details were removed. Anything in angle brackets is a placeholder for your own setup.