From 5490141cf616ad7e39b55d4983a4ac1969f81fbf Mon Sep 17 00:00:00 2001 From: Clawdbot Date: Tue, 24 Mar 2026 21:00:11 +0000 Subject: [PATCH] Initial commit: Ignition skill for viral X/Twitter content --- SKILL.md | 106 ++++++++++++++++++++++++++++++++++++++++++ references/formats.md | 75 ++++++++++++++++++++++++++++++ references/hooks.md | 52 +++++++++++++++++++++ 3 files changed, 233 insertions(+) create mode 100644 SKILL.md create mode 100644 references/formats.md create mode 100644 references/hooks.md diff --git a/SKILL.md b/SKILL.md new file mode 100644 index 0000000..24844fd --- /dev/null +++ b/SKILL.md @@ -0,0 +1,106 @@ +--- +name: ignition +description: | + Viral content creation for X/Twitter using the Ignition method (Catalyst + Winning Format + Content Matrix). + Adapted from The School of Hard Knocks' 140M views/month framework for text-based platforms. + Use when: creating viral tweets, threads, or content series for X/Twitter. + Transforms a topic into high-retention content by finding the right hook type, locking a winning format, and ensuring replicability at scale. + NOT for: long-form articles, YouTube videos, or platforms where visual format dominates over text. +--- + +# Ignition 🔥 + +140 million views in one month. 1 million followers in 30 days. The method behind it is stupidly simple — and that's why it works. + +Adapted from The School of Hard Knocks framework for X/Twitter text content. + +## The Three Pillars + +### 1. The Catalyst + +Two components: **Constants** and **Hooks**. + +**Constant** = your fixed audience. Pick one and never waver. Every piece of content speaks to that person. + +On X/Twitter, the Constant is baked into your bio and reinforced by every tweet. If your audience is founders, every tweet assumes the reader is a founder. No explaining, no broadening. Narrowing is what makes hooks land. + +**Hooks** = the first line of the tweet. Three types that work: + +| Type | What it does | X/Twitter example | +|------|-------------|-------------------| +| **Strong claim** | Says something most people disagree with or find surprising | "Kubernetes é overkill pra 99% das startups. Eu uso pra 100% das minhas." | +| **Direct approach** | Addresses the reader's situation head-on | "Você tá rodando 3 agentes de IA em paralelo e o throughput caiu. Eu sei por quê." | +| **Simple question** | Triggers curiosity with a question they can't ignore | "Quantas horas por semana você gasta revisando código que um agente escreveu?" | + +**The Catalyst rule:** Test all three. When one type consistently outperforms, repeat it relentlessly. Don't rotate for variety — rotate for data, then lock in the winner. + +See [references/hooks.md](references/hooks.md) for hook formulas and examples. + +### 2. Winning Format + +The format that retains attention on X/Twitter threads: + +**The Pullaway** — Start with something that feels like you're about to walk away from the topic. The reader stays because it seems like something important is about to drop. + +On X/Twitter this translates to: +- T1: Hook + setup that creates tension (the pullaway) +- T2-Tn: Question → Answer → Question → Answer chain +- Each tweet answers the previous tension and opens a new one + +**Question stacking** is the engine. Every answer births the next question. The reader never reaches a resting point. + +Applied to threads: +``` +T1: [Hook + pullaway — tension created] +T2: [Answers T1's tension, raises new question] +T3: [Answers T2, introduces complication] +T4: [Resolves complication, opens deeper layer] +... +Tn: [Payoff + CTA] +``` + +See [references/formats.md](references/formats.md) for format templates. + +### 3. Content Matrix + +Every piece of content must pass three gates simultaneously: + +| Gate | Question | Fail = don't publish | +|------|----------|---------------------| +| **Easy to make** | Can I produce this in under 30 minutes? | If it takes 2 hours, it won't scale | +| **Replicable** | Can I make 10 more with the same structure? | If it's a one-off insight, it's not a format | +| **Engageable** | Will someone reply, quote, or save this? | If it only gets likes, it won't spread | + +The matrix is a filter, not a goal. Content that passes all three gates gets published. Content that fails any gate gets reworked or killed. + +**Replicability is the multiplier.** A format you can reuse 50 times with different topics beats a brilliant one-off every time. + +## Process + +### Input required +1. **Topic/niche** — What to create content about +2. **Constant** — Who is the fixed audience (default: from profile/bio) +3. **Hook type to test** — Strong claim, direct approach, or simple question (or test all three) +4. **Format** — Single tweet, thread, or series (default: thread) +5. **Language** — PT-BR or EN (default: PT-BR) + +### Output + +For each piece: +1. **Hook** (first tweet) — Must pass the 1-second test: would you stop scrolling? +2. **Body** — Question-answer chain with pullaway tension +3. **CTA** — Clear, specific, single action +4. **Matrix check** — Confirm all three gates pass +5. **Replication template** — The structure with blanks, ready for next topic + +### Combining with other skills + +- **Sherazade**: Use Ignition for individual thread optimization, Sherazade for multi-day series sequencing. Ignition handles the micro (each tweet retains), Sherazade handles the macro (each thread pulls to the next). +- **Humanizer**: Always run Humanizer as final pass. Ignition formats can drift toward formulaic if not humanized. + +## Anti-patterns + +- **Rotating hooks for variety** — Find what works, then repeat. Variety is the enemy of data. +- **Complex formats** — If you can't explain the format in one sentence, it's too complex to replicate. +- **Content that only you can make** — If it requires unique insider knowledge every time, it fails the replicability gate. +- **Optimizing before testing** — Publish 10 variations before optimizing 1. diff --git a/references/formats.md b/references/formats.md new file mode 100644 index 0000000..3ed7b11 --- /dev/null +++ b/references/formats.md @@ -0,0 +1,75 @@ +# Winning Formats for X/Twitter + +## The Pullaway Thread + +The core format. T1 creates tension by seeming like you're walking away from the insight. Reader stays because payoff feels imminent. + +``` +T1: [Hook — strong claim or situation] + [Pullaway — "E aí eu descobri..." / "Levei X meses pra entender..." / Setup incompleto] + 🧵 + +T2: [Context — what was happening before] + [First question planted implicitly] + +T3: [Answer to T2's question] + [New complication introduced] + +T4: [Resolution of complication] + [Deeper layer revealed — "mas o problema real era..."] + +T5: [Core insight delivered] + [Connects back to T1's hook] + +T6: [Practical takeaway + CTA] + [Cliffhanger for next thread if part of series] +``` + +### Rules +- Each tweet must make sense alone (someone might land mid-thread) +- No tweet should fully resolve the tension — always open the next loop +- T1 must work as a standalone tweet (it's what gets retweeted) +- Last tweet carries the CTA — save, follow, or link + +## Question-Answer Chain (Single Tweets) + +For standalone tweets, compress the pullaway into a single unit: + +``` +[Hook — question or claim] +[2-3 lines of answer that opens a bigger question] +[Punchline that reframes everything] +``` + +**Example:** +"Quantos PRs você revisa por dia? + +Eu revisava 3-4. Levava 30 minutos cada. Agora o agente manda screenshot de cada tela, resumo do que mudou, e o que pode quebrar. + +Rejeito pelo print. Nem logo no sistema. Review caiu pra 8 minutos." + +## The Contrast Format + +Two states side by side. Before/after, wrong/right, expectation/reality. + +``` +[Estado 1 — como era / como a maioria faz] + +[Estado 2 — como ficou / como funciona] + +[Uma frase conectando os dois] +``` + +Works well for strong claims. The contrast does the persuading. + +## Matrix Check Template + +Before publishing, verify: + +| Gate | ✅ / ❌ | Notes | +|------|--------|-------| +| Easy to make (<30 min) | | | +| Replicable (same structure, different topic) | | | +| Engageable (reply/quote/save trigger) | | | + +All three must pass. Two out of three = rework. diff --git a/references/hooks.md b/references/hooks.md new file mode 100644 index 0000000..0c783ad --- /dev/null +++ b/references/hooks.md @@ -0,0 +1,52 @@ +# Hook Formulas for X/Twitter + +## Strong Claim + +The reader disagrees or is surprised. They stop to see if you back it up. + +**Formulas:** +- "[Coisa popular] é [opinião contrária]. [Prova curta]." +- "Todo mundo fala pra fazer [X]. Eu fiz o oposto e [resultado]." +- "[Número absurdo] em [tempo curto]. [Setup]." +- "Eu [ação polêmica] e [resultado inesperado]." + +**Examples (PT-BR):** +- "Meu agente de IA deletou o próprio binário semana passada. Pod reiniciando a cada 30 segundos." +- "Parei de fazer code review e o throughput triplicou." +- "3 agentes em paralelo é pior que 1. Goldratt explicou por quê há 40 anos." + +## Direct Approach + +You describe the reader's exact situation. They feel seen. + +**Formulas:** +- "Você tá [situação específica]. [Eu sei por quê / Eu passei por isso]." +- "Se você [condição], [consequência que eles já sentem]." +- "[Número] de vocês vão [ação]. [Número menor] vão [resultado]." + +**Examples (PT-BR):** +- "Você tá botando agente de IA pra codar e o backlog tá maior que antes. Normal." +- "Se você roda CI sem staging isolado, cada merge é roleta russa." +- "Você automatizou o desenvolvimento e agora o gargalo é você. Eu sei porque aconteceu comigo." + +## Simple Question + +Triggers curiosity. They answer mentally and need to see if they're right. + +**Formulas:** +- "Quantas [métrica] você [ação] por [período]?" +- "O que acontece quando [cenário hipotético]?" +- "Por que [coisa que todo mundo faz] [não funciona como esperado]?" + +**Examples (PT-BR):** +- "Quantas horas por semana você gasta resolvendo conflito de merge?" +- "O que acontece quando seu agente de IA produz mais rápido do que você revisa?" +- "Por que botar mais dev no projeto não acelera a entrega?" + +## Testing Protocol + +1. Write the same content with all three hook types +2. Publish at similar times on different days +3. Track: impressions, replies, bookmarks (not just likes) +4. After 10 tests, lock the winner +5. Use the winner for the next 30 posts minimum before re-testing