---

name: substack-write description: Write a Substack post in Anuj's voice — sharp/contrarian edge, 600-1200 words, email-first cadence, subject line and dek, no engagement-bait. Outputs Markdown ready to paste into Substack.

substack-write

Drafting a Substack post for Anuj Sadani. Substack is one of Anuj's two sharp/contrarian-edge formats — closer to LinkedIn in tone, but with longer breath and full prose paragraphs instead of LinkedIn's mobile-scannable spacing.

Step 1 — Load context (mandatory)

Read these files:

1. shared/voice-rules.md
2. shared/voice-samples.md
3. shared/pet-peeves.md
4. shared/topic-modes.md
5. shared/platform-styles.md

Step 2 — Classify topic mode + apply platform style profile

Apply topic-modes.md.

Then apply the substack row from shared/platform-styles.md: technical depth 3 (one short explained snippet only if the topic needs it), headline aggressiveness 4 (intriguing subject for open-rate, never engagement-bait), density 3 (medium prose, one pull-quote). Honor the precedence rules in that file — voice-rules.md/pet-peeves.md are absolute, and topic mode can raise (never lower) substance.

Step 3 — Plan the structure

Default outline (600-1200 words, 4-6 min read):

1. Subject line — what shows up in the inbox. Punchy, specific, ≤8 words. Echoes the hook of the post body.
2. Dek — 1 sentence under the subject. Sets up the tension.
3. Opening — a hook line in one of the cadences from voice-samples.md (dichotomy, demoting, pragmatic-flex, observation-flip).
4. 3-5 short sections — each is 2-4 paragraphs. Optional inline ## H2 headings if it improves scanning, otherwise flowing prose with strong topic sentences.
5. One concrete example or number per section.
6. Closer — 2-3 lines. Crisp takeaway OR one genuine open question. Never engagement-bait.

Step 4 — Draft

Tone: closer to LinkedIn-sharp than Medium-measured. Owned opinions. First person. Email-first: assume the reader is scanning in their inbox.

Emoji: follow the Emoji Control section in shared/voice-rules.md. Default level is low. Override only if the user explicitly asks.

• Optional pull-quote (Markdown blockquote > ...) once per post for emphasis.
• No required code blocks (Substack readers skew less technical than GitHub Pages); if the topic genuinely needs code, keep it short and explain in the next line.
• No images required.

Step 5 — Pre-delivery checks

1. Run the regex sweep from shared/pet-peeves.md. Any hit → rewrite.
2. Confirm word count is in the 600-1200 range.
3. Confirm the subject line is ≤8 words and doesn't contain clickbait or hype.
4. Confirm closer is not engagement-bait.
5. Confirm: style profile applied — depth, headline, and density match the substack row in shared/platform-styles.md (topic-mode overrides may raise depth).

Step 6 — Save and deliver

Save to file. Derive a kebab-case topic slug. Get today's date in YYYY-MM-DD form. Save subject + dek + body to:

./drafts/<YYYY-MM-DD>-<topic-slug>/substack.md

Format the file as:

# Subject: <subject line>
Dek: <dek>

---

<body in Markdown>

If the file already exists, suffix with -v2, etc. Create directories as needed.

Also print in chat:

─── SUBJECT ───
<subject line, ≤8 words>

─── DEK ───
<one sentence>

─── POST ───
<body in Markdown>

─── META ───
words: <N>
reading time: <N> min
topic mode: <security | architecture-agents | leadership | cost-infra>
emoji level: <none | low | medium | high>
saved to: <path>

User copies to Substack manually.