jab-hook
Gary Vaynerchuk's jab-jab-jab-right-hook framework applied to a personal portfolio rotation on X and LinkedIn.
From Corey Haines' makerskills, MIT, pinned at cc31579. Updated 2026-10-08.
The file, as Kwern runs it
Nothing below is rewritten or cut. The description a model routes on comes first, then the file itself. The whole thing, frontmatter included, is at /skills/jab-hook.md.
description
Gary Vaynerchuk's jab-jab-jab-right-hook framework applied to a personal portfolio rotation on X and LinkedIn. Jabs = build-in-public + educational (value). Hooks = promo (the ask). Each property in the user's configured portfolio (see `~/.config/makerskills/jab-hook/properties.yaml`) gets a hook at least once every ~3 weeks; jabs fill the rest. Drafts go into the user's Typefully workspace via MCP. Modes — plan (7-day plan), pick-next (single post), audit (coverage report), draft (specific post). Triggers on "/jab-hook," "what should I post," "plan my socials," "next promo," "next jab," "next hook," "social rotation," "promote [property]," "BIP post," "audit my socials," "what haven't I posted about."
Gary Vaynerchuk's framework applied to your configured portfolio: keep the feed mostly jabs (value — BIP + educational) so the hooks (promo) earn attention when they land. Drafts to your personal Typefully workspace.
One-time setup: see references/properties.md and references/typefully-config.md to configure your real portfolio + workspace. Personal config lives in ${MAKERSKILLS_CONFIG:-$HOME/.config/makerskills}/jab-hook/.
Mental model
- N rotation slots (configured in
~/.config/makerskills/jab-hook/properties.yaml): one per property you want to promote on rotation - ~2 promo posts/week → each property cycles every ~3 weeks at equal weight
- Non-promo days: 50/50 build-in-public vs educational
- Target weekly rhythm: ~2 promo + ~2–3 BIP + ~2–3 educational
- Platforms: X + LinkedIn via your personal Typefully workspace
Step 1 — Pick the mode
Detect from the trigger:
| User says | Mode |
|---|---|
| "plan my socials," "next week's posts" | plan |
| "what should I post," "next promo," "pick next" | pick-next |
| "audit my socials," "what's overdue," "what haven't I posted about" | audit |
| "draft a promo for X," "BIP post about Y," "educational about Z" | draft |
If ambiguous, confirm.
Step 2 — Load context
- Read
references/properties.md— the 6 rotation slots and angle ideas - Read
references/voice.md— the user's voice rules per platform - Read
references/content-types.md— templates for promo / BIP / educational - Pull recent posts from your personal Typefully workspace:
- First run: call the Typefully MCP server's list-social-sets tool and ask which social set is your personal (X + LinkedIn). Save the ID to
${MAKERSKILLS_CONFIG:-$HOME/.config/makerskills}/jab-hook/typefully.yamlfor future runs (schema inreferences/typefully-config.example.yaml— never save it inside the skill folder; upgrades wipe it). - Call its list-drafts tool, filtered to the last 30 days
- Classify each as promo / BIP / educational by content
- First run: call the Typefully MCP server's list-social-sets tool and ask which social set is your personal (X + LinkedIn). Save the ID to
- Compute days since last promo for each of the 6 slots
Step 3 — Mode logic
plan (7-day plan)
- Slot the 1–2 most overdue properties as promo posts
- Fill remaining 5 days with alternating BIP and educational (50/50)
- Output as a table:
Day | Type | Property/Topic | Hook | Draft summary - Don't stack two promo on consecutive days
- After the user approves the plan, offer to draft each one in sequence
pick-next
- Identify the most overdue property
- If most overdue < 14 days, suggest a BIP or educational instead (whichever is less recent)
- Draft 1 post; ask before syndicating
audit
- Report: days since last promo per property, ordered by most overdue
- Flag any property >21 days as overdue
- Suggest the next 1–2 moves
draft
- Skip rotation logic — draft the requested post
- Use the right content-type template
Step 4 — Draft
Follow references/voice.md and references/content-types.md. Default: single post, one per platform (X version + LinkedIn version). Thread only if the second post earns its place.
Link placement (load-bearing): never put a URL in the body. Default is no link at all — let curiosity drive the click. When a link is needed, draft it as a first comment for LinkedIn (Typefully supports auto-comment) or a reply for X. The body always stands alone.
Optional: inspiration scan
For promo and educational posts, optionally scan inspiration accounts (references/inspiration.md) before drafting:
- Pick 1–2 accounts whose audience overlaps with the property
- Pull recent posts (agent-browser if it works on LinkedIn; otherwise ask the user to paste examples or fall back to their X/newsletter)
- Extract structural patterns: hook openers, post length, line-break rhythm, CTA styles
- Apply ONE pattern to the user's voice — never mimic phrasing
- Log new patterns in
${MAKERSKILLS_CONFIG:-$HOME/.config/makerskills}/jab-hook/patterns.local.md(create if missing) so they accumulate —references/patterns.mddocuments the starter patterns and stays read-only
Skip inspiration scan when the user says "just draft it" or when iterating on an existing post.
Step 5 — Syndicate to Typefully
Ask: "Push to Typefully now? (X + LinkedIn, your personal workspace)"
If yes:
- Read social set ID from
${MAKERSKILLS_CONFIG:-$HOME/.config/makerskills}/jab-hook/typefully.yaml(setup notes inreferences/typefully-config.md) - Call the Typefully create-draft tool once for X, once for LinkedIn (or once with both platforms if the social set spans both)
- If the draft has a first-comment link, configure Typefully to add it automatically (LinkedIn auto-comment / X reply). If the MCP doesn't support auto-comment for one of the platforms, surface the comment text so the user can paste it after publishing.
- Default to draft state (not scheduled) — the user reviews in Typefully UI before sending
- Return the Typefully draft URLs
Composes with
paste— clean output for pasting into Typefully / LinkedIn / X manually if the MCP path doesn't fitsocial-fetch— pull inspiration-account posts for structural analysis (replaces inline agent-browser)second-brain— theContent Ideaswiki page hoppers hooks, frameworks, stories;/jab-hookdrafts pull candidates from theredeep-research— when a promo needs a stat or citation the draft doesn't have yetmarketingskills:social— generic social frameworks, useful when teaching strategy vs shipping your own postsmarketingskills:copywriting— for hook / headline ideation when stuck
Notes on quality
- 1 promo per week per property is the floor, not the ceiling. Rotation compounds when it's predictable — 4 properties × 1 promo/week = 4 promo weeks/month. Educational + BIP + community fill the rest.
- Never mimic phrasing from inspiration accounts. Extract structure (hook opener type, line-break rhythm, CTA style) and apply to your own voice. Voice is the moat; copying phrasing destroys it.
- Links belong in first comments, not the body for LinkedIn + X. Documented in
feedback_social_link_placement.mdin your memory store (see below). Body posts with inline URLs get algorithm-suppressed on both platforms. - Draft state, not scheduled state. Every Typefully push defaults to draft — human review in the Typefully UI before publish is non-negotiable. Automated scheduling of unreviewed drafts has repeatedly produced posts the user regrets.
- Reader-perspective framing. "You'll know X" not "we'll teach you X." "What you're guaranteed" not "what we'd guarantee."
- Voice.local.md overrides the shipped voice.md. Personal voice rules never live in the public repo — always in
${MAKERSKILLS_CONFIG:-$HOME/.config/makerskills}/jab-hook/voice.local.md. - Cadence cap: 2 posts/day per platform, per account. More than that degrades reach + reads as spam.
Memory references
Your memory store is your agent's persistent memory if it has one (Claude Code's memory directory, for example), otherwise ${MAKERSKILLS_MEMORY:-${MAKERSKILLS_CONFIG:-$HOME/.config/makerskills}/memory}/.
feedback_promo_voice.mdin your memory store (if present) — conviction-coded CTAs + reader-perspective framingfeedback_social_cadence.mdin your memory store (if present) — cadence cap (e.g., 2 posts/day per platform)
What it discloses
The Skill reads these files only when it needs them, so the model spends nothing on them the rest of the time.
From your accounts to a proposal
- 01
Connect what it reads
Give the X posts agent the accounts it works on. Kwern never asks for a password, and an agent missing a connection says so instead of guessing.
- 02
The loop runs every 3 days
Before it writes a word it reads the foundation doc, which holds the positioning, the audience and the voice, and whatever its connectors return that day.
- 03
The jab-hook skill runs
The agent activates it when the run calls for it, and what the model then reads is the file on this page, word for word.
- 04
A proposal waits for you
Nothing is published and nothing is spent until you approve it. Approve and Kwern posts it where it holds the keys; where it does not, it hands you the text and the link.
The file, the agency, and the agent
| On your own | An agency | Kwern | |
|---|---|---|---|
| The method | Whatever you remember to do that week | Theirs, and you never get to read it | This file, open on this page |
| Who runs it | You, when there is time | An account manager, between other accounts | The X posts agent, every 3 days |
| What it reads | What you can pull by hand | What you remembered to send them | Your connected accounts and your foundation doc |
| What comes out | A draft, sometimes | A deck, and a report at the end of the month | A proposal you approve or throw away |
| What it costs | Your evenings | A retainer | One subscription, and the model on your own key |
Questions about jab-hook
Pair it with
The other skills the same agents reach for during a run.
Someone is asking for what you built today.
On Reddit, on X, in ChatGPT's answers. Kwern finds them every morning and drafts the reply in your voice; you decide what goes out.