notslop-write-blog-post

name	notslop-write-blog-post
description	Use when the user wants to write a long-form blog post (1500-3000 words) on a topic. Pulls a week of social signal, clusters it into themes, identifies undercovered angles, then produces an SEO-aware article with per-section word counts, E-E-A-T citations of real posts, and citation-ready structure for AI Overviews.

Providers required

Capability	Required	Providers (BYOK)	Setup	Cost
Pull social signal (Reddit/HN/blogs)	yes	built-in, no key needed	—	free
Rerank by relevance	yes	ZeroEntropy zerank-2	PROVIDERS.md#zeroentropy-rerank--embed	free tier OK
Embed for theme clusters + dedup	yes	ZeroEntropy zembed-1	PROVIDERS.md#zeroentropy-rerank--embed	free tier OK
Scrape X posts	only if topic needs X data	Orthogonal (ScrapeCreators)	PROVIDERS.md#x-twitter--via-orthogonal	~$0.02/handle, $10 free at signup

When to use this skill

• "Write me a blog post about X"
• "Draft a long-form article on Y for my site"
• "I need a 2000-word piece on Z"
• The user mentions blog, article, long-form, or asks for word counts >1000.

For X long-form Articles (Notes format), use notslop-write-x-article. For just the title + meta description, use notslop-write-blog-headline. For LinkedIn long-form, use notslop-write-linkedin-post.

Setup

notslop init configured. ZeroEntropy key set.

Steps

1. Gather inputs from the user (ask only for what's missing):

   • Topic / focus keyword (the primary SEO target)
   • Target audience (e.g. AI builders, growth marketers, devs)
   • Target length (default 2000 words; valid range 1500-3000)
   • Voice/tone (analytical / opinionated / tutorial)
   • Author angle: their experience or unique POV on the topic
   • Optional: their website / publication for tone reference

2. Research phase. Run the four commands. Pipe each to a temp file or read sequentially; do not skip any:

   notslop digest "<TOPIC>" --since 7d --top 15 --for-content
   

   notslop pulse "<TOPIC>" --window 7d --top 5
   

   notslop voices "<TOPIC>" --top 10 --for-content
   

   notslop find-related "<TOPIC> — <USER ANGLE>" --top 10
   

3. Synthesize the research before drafting. In your own working notes, identify:

   • The 3-5 dominant themes from pulse
   • 5-10 quotable real posts with URL + date + author from digest and voices
   • 2-3 undercovered angles that the recent signal does not address
   • The most cited claim or framing the user could reinforce or push back on

4. Produce the content brief BEFORE the draft. Output a markdown block titled ## Content brief containing:

   • Focus keyword (1 primary) + 2-3 secondary keywords
   • H1 (the title — see step 5 for rules)
   • Meta description (155-160 chars, focus keyword in the first 100 chars)
   • Outline: 4-7 H2 sections, each with:
     • Section heading
     • 2-3 bullet sub-points it must cover
     • Word budget (sum of all sections within ±10% of target length)
     • Which real posts to cite in this section (URL + author)
   • Estimated reading time (words / 220 wpm, rounded to nearest minute)

5. Title rules (H1 + meta):

   • 50-70 chars (hard cap 70)
   • Front-loaded with focus keyword
   • Specific outcome or claim, not "Guide to X" / "Everything you need to know about X"
   • No "10 ways", no "ultimate guide", no clickbait inflation
   • Pattern that works for AI/dev audiences: <Specific outcome>: <how/why> <focus keyword>

6. Write the article. Structure rules:

   • Open with a specific, dated observation from the social signal (gives E-E-A-T signal in the first paragraph). Example: "On May 8, @author posted that..."
   • Each H2 section follows claim → evidence (cite real post with URL) → implication. 200-400 words depending on the brief.
   • Use blockquotes for direct citations from posts.
   • Include a ## Citations section at the bottom with all referenced posts (numbered, URL, author, date).
   • Closing: a falsifiable prediction or a specific question for the reader. Never "What do you think?".

7. Readability + SEO checks. Before showing the draft, verify and self-report:

   • Avg sentence length <25 words. Flag any paragraph that exceeds 100 words.
   • Headings every 200-300 words.
   • Each H2 is self-contained (an AI Overview could quote it without surrounding context). Use clear claim + evidence pattern inside each section.
   • Focus keyword density 0.5-1.5%. Appears in title, H1, first 100 words, 2-3 H2s, meta description.
   • Every non-first-person claim cites a real post URL from the digest.

8. AI-slop scan. Search the draft for these phrases. If any appear, flag them in a separate block for user review (do NOT silently delete):

   • "rapidly evolving landscape"
   • "in today's world"
   • "the future of X"
   • "leverage"
   • "robust"
   • "seamlessly"
   • "delve into"
   • "navigate the complexities"
   • "in conclusion"

Output format

Output in this exact order:

1. The content brief (markdown, the full block from step 4).
2. The full article (markdown).
3. The AI-slop scan results — either "No flagged phrases found." or a numbered list of phrases + line context.
4. A 1-tweet teaser (≤ 280 chars) the user can post on X to drive traffic to the article.

Constraints

• No "in today's rapidly evolving landscape" or any flagged AI-slop phrase in the final draft (flag, don't auto-edit; the user decides).
• Every claim that isn't first-person experience must cite a real post URL from the digest. No inventing sources.
• Avg sentence length <25 words. Flag any paragraph >100 words.
• Reading time displayed once at the top of the article.
• If the user explicitly asks for an opinion piece (no citations), keep E-E-A-T via first-person experience markers ("I ran into this when...", "We shipped X and saw Y") and skip the ## Citations section.
• Word count must land within ±10% of the user's target.
• No emojis unless the user asks.
• No hashtags.