Addictive writing
Drafts, rewrites, and reviews prose and scripts for earned attention using clear stakes, question-driven momentum, causal progression, expectation-and-reveal, re-hooks, rhythm, conversational tone, distinctive story lenses, purposeful endings, and medium-appropriate hooks. Use when the user asks to make writing more compelling, gripping, page-turning, bingeable, high-retention, suspenseful, or engaging; diagnose hooks, pacing, narrative tension, story structure, transitions, or reader drop-off; or improve fiction, nonfiction, articles, essays, newsletters, speeches, presentations, sales narratives, video scripts, podcasts, and social content without sacrificing truthfulness or the author's voice.From its SKILL.md
npx -y skills add MarcoWorms/addictive-writingAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its file declares
Copied from the file, not written here
The file declares its own license as Apache-2.0. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.
SKILL.md
23.8 KB, ~4.9k tokens by cl100k_base, as published. Nobody here has run it
Addictive Writing
Create earned attention: writing that readers continue because each passage creates meaningful anticipation and then rewards it.
Do not interpret “addictive” as permission to deceive, exploit, sensationalize, or obstruct informed choice. Optimize for curiosity, clarity, emotional relevance, and satisfying payoffs.
Operating priorities
Apply these priorities in order:
- Truth and safety — preserve facts, uncertainty, attribution, and material qualifications.
- Reader contract — fulfill the promise made by the title, hook, premise, or assignment.
- Meaning and voice — preserve the author’s intent, point of view, genre, and level of formality.
- Causal clarity — make events and ideas feel connected rather than merely adjacent.
- Attention — create stakes, prediction, contrast, revelation, and forward motion.
- Style — improve rhythm, tone, imagery, specificity, and memorability.
If a higher priority conflicts with a lower one, protect the higher priority.
Non-negotiable rules
- Never invent facts, quotations, sources, results, motives, testimonials, scarcity, deadlines, or stakes.
- Never hide information that a reader needs for a medical, legal, financial, safety, or other high-stakes decision.
- Never overstate in either direction: preserve uncertainty, and do not turn incomplete evidence into “impossible,” “no effect,” or another absolute unless the source supports it.
- Never promise a payoff the piece does not deliver.
- Never use a random twist when an earned reframe is possible.
- Never make every sentence dramatic. Contrast requires calm, orientation, and release.
- Never flatten a distinctive voice into generic creator-speak, sales copy, or clickbait.
- Never claim that a line “causes dopamine,” “hacks the brain,” or makes disengagement chemically impossible.
- Prefer specific, benchmarked relevance over vague hype or ungrounded labels such as “heavy,” “significant,” or “practical.”
- Prefer causality over chronology.
- Prefer an earned surprise over an arbitrary shock.
- Prefer a satisfying answer over endless teasing.
The core attention loop
Use this loop at the scale appropriate to the piece:
Stakes → Big question → Escalation → Turn → Payoff → Re-hook
1. Stakes
Give the reader a reason to care before asking them to wait for an answer.
Establish as many of these as the genre needs:
- Character or subject: Who or what are we following?
- Desire or objective: What outcome is wanted?
- Risk or cost: What changes if this fails, remains unknown, or goes wrong?
- Urgency: Why does it matter now rather than eventually?
- Relevance: Why should this specific audience care?
Stakes may be emotional, practical, intellectual, reputational, moral, financial, relational, or physical. They do not need to be life-or-death.
2. Big question
Open a specific, meaningful uncertainty the reader can mentally work on.
A strong question is:
- concrete enough to invite a prediction;
- important enough to justify attention;
- bounded enough to be answerable;
- connected to the central promise;
- introduced early enough to generate momentum.
The question may be explicit or implicit. Avoid empty teasers such as “something crazy happened” or “you will not believe this.”
3. Escalation
Complicate the path to the answer without wandering away from it.
Escalate through one or more of these:
- a stronger obstacle;
- a narrowing deadline;
- a higher cost of failure;
- new evidence that changes the likely answer;
- a forced choice or tradeoff;
- a revelation that deepens the original problem;
- a reversal in power, knowledge, or intention.
Escalation must change the reader’s model of the situation. More detail alone is not escalation.
4. Turn
Break the most likely prediction with a reveal that is surprising before it lands and coherent after it lands.
Build an earned turn by:
- identifying what a reasonable reader will predict;
- planting fair clues or constraints;
- revealing a contrasting answer;
- making the earlier clues newly meaningful;
- avoiding information that exists only to mislead.
A good turn produces “I did not see that coming, but it makes sense.” A bad turn produces “Where did that come from?”
5. Payoff
Answer the question, complete the promised demonstration, or deliver the emotional or practical value the setup created.
A payoff should be:
- proportional to the setup;
- clear enough to recognize;
- specific rather than abstract;
- useful, moving, funny, revealing, or consequential;
- followed by enough space for the reader to register it.
Do not confuse a re-hook with a payoff. Close the current loop before relying on the next one.
6. Re-hook
Open the next meaningful uncertainty before momentum falls to zero.
Useful re-hook shapes include:
- “That solved the first problem—but created a worse one.”
- “The test worked, which meant the original theory was wrong.”
- “We found the missing file. It had been created tomorrow.”
- “She agreed to the deal—except the signature was not hers.”
- “The answer was simple. Applying it was not.”
A re-hook is connective tissue, not another vague promise.
The context–conflict dance
Alternate orientation with change:
- Context: establish the current goal, state, evidence, or plan.
- Conflict: introduce an obstacle, contradiction, exception, cost, or unanswered implication.
- Consequence: show what the conflict causes.
- Response: show the next choice, action, inference, or adjustment.
- Repeat only while each beat materially changes the situation.
Use causal and contrastive links such as:
- because
- therefore
- so
- but
- yet
- until
- except
- which meant
- as a result
- despite that
Treat repeated “and then” connections as a diagnostic warning. Replace chronology with cause, consequence, contrast, or intentional omission where truthful.
Task routing
Determine the mode before working:
- Draft: create a new piece from a brief, outline, evidence set, or premise.
- Rewrite: improve an existing piece while preserving meaning and voice.
- Review: diagnose strengths, attention leaks, truth risks, and revision priorities.
- Outline: design the causal spine and loop structure before prose.
- Punch-up: improve selected hooks, transitions, turns, endings, or passages without rebuilding the whole piece.
Use the lightest mode that satisfies the request. Do not force a full structural rewrite when line editing is enough.
Default workflow
Step 1: Establish the reader contract
Identify or infer:
- target audience;
- medium and expected length;
- desired reader action, feeling, or understanding;
- central promise;
- source material and factual boundaries;
- voice, genre, and intensity;
- required ending or call to action.
When information is missing, make conservative assumptions and preserve easy reversibility. Do not fabricate specificity.
Step 2: Protect the source of truth
Before changing factual material, extract:
- verified facts and figures;
- direct quotations;
- attributed claims;
- uncertainty and caveats;
- chronology that must remain accurate;
- claims that need support;
- details that must not be altered.
For fiction, replace “facts” with canon, characterization, world rules, point of view, and continuity constraints.
Step 3: Define the destination first
Write a one-sentence destination:
By the end, the reader should understand, feel, or decide [specific outcome].
Then draft a candidate final line or final beat. The ending should resolve the central promise, create an aftereffect, and avoid draining the piece with redundant summary.
For looping short-form content, the last line may naturally reconnect to the opening. Do not create an awkward loop merely to force replay.
Step 4: Choose a story lens
Find the most useful and distinctive angle, not merely the broad topic.
Generate up to three lenses using combinations such as:
- familiar subject × overlooked cause;
- familiar event × unexpected consequence;
- accepted belief × disconfirming evidence;
- visible result × hidden mechanism;
- public story × private decision;
- success × cost;
- failure × useful lesson;
- trend × second-order effect.
Choose the lens with the best combination of audience relevance, evidence, novelty, and fit with the author’s authority. Do not choose novelty unsupported by the material.
Step 5: Build the causal spine
Reduce the piece to beats. For each beat, record internally:
- what the reader knows;
- what the reader currently wants to know;
- what the reader is likely to predict;
- what changes;
- why the next beat follows;
- which loop closes;
- which loop opens.
Use this compact beat pattern when useful:
Because [cause], [action or consequence]. But [complication]. Therefore [next action or inference].
Remove beats that neither change the situation nor improve necessary understanding.
Step 6: Design the opening
Make the first movement do real work. Select the opening type that best fits the material:
- Concrete anomaly: a specific fact or image that contradicts expectation.
- Goal under pressure: a person wants something and a meaningful obstacle is already present.
- Result plus mystery: reveal the outcome, then ask how it happened.
- Sharp contrast: place the common belief beside the less obvious reality.
- Useful promise: state exactly what the reader will learn, see, or be able to do.
- Decision point: begin at the irreversible choice, then supply context.
The opening should indicate the actual subject. Do not use an opaque hook that could introduce any topic.
For visual media, pair the opening claim with a visible object, result, demonstration, contrast, document, environment, or action that confirms the premise.
Step 7: Draft in attention loops
Draft the minimum context needed to understand the next change. Then alternate context, conflict, consequence, and response.
At major reveals:
- give enough information for a reasonable prediction;
- delay only as long as the delay increases understanding or tension;
- deliver the answer cleanly;
- let it register;
- re-hook through a consequence, complication, or deeper question.
Use nested loops sparingly:
- one central loop carries the whole piece;
- section-level loops carry major movements;
- paragraph- or scene-level loops maintain local momentum.
Do not leave more open loops than the reader can remember.
Step 8: Improve tone and rhythm
Write to one intelligent person, not to a demographic abstraction.
- Prefer direct, natural syntax.
- Use contractions when they fit the voice.
- Use questions only when the reader can genuinely answer or predict.
- Replace throat-clearing with subject matter.
- Prefer concrete nouns and active verbs.
- Preserve technical language when precision requires it; define it at the moment of need.
- Avoid faux intimacy, forced slang, and repetitive creator catchphrases.
Vary sentence length by function:
- short sentences land decisions, shocks, conclusions, or emphasis;
- medium sentences carry most explanation and action;
- long sentences build accumulation, complexity, momentum, or reflection.
Do not mechanically alternate lengths. Read aloud and revise monotony, breathlessness, accidental tongue-twisters, and emphasis that lands on the wrong word.
For scripts, place spoken sentences or meaningful beats on separate lines when that helps reveal cadence and visual pacing.
Step 9: Strengthen transitions
At every section boundary, ask whether the reader has permission to stop.
Replace dead transitions such as “Next, let us discuss…” with consequence-driven movement:
- implication;
- contradiction;
- unresolved cost;
- changed objective;
- new evidence;
- a narrower question;
- a decision forced by the previous answer.
A transition should explain why the next section belongs now.
Step 10: Verify and revise
Run these passes in order:
- Truth pass: facts, attribution, caveats, chronology, and uncertainty remain intact.
- Promise pass: the opening promise and central question receive clear payoffs.
- Causality pass: each major beat follows through cause, consequence, contrast, or purposeful juxtaposition.
- Prediction pass: the reader has enough information to form expectations before important reveals.
- Turn pass: surprises are prepared, coherent, and non-deceptive.
- Re-hook pass: momentum does not die immediately after major payoffs.
- Compression pass: remove setup, repetition, and explanation that no longer earns its space.
- Voice pass: the result still sounds like the intended author and genre.
- Rhythm pass: read aloud or simulate oral delivery.
- Ending pass: stop on the strongest complete note.
Revise the two weakest dimensions before polishing everything else.
Medium-specific guidance
Fiction and narrative nonfiction
- Let desire, obstacles, choices, and consequences carry the loop.
- Preserve ambiguity when it is purposeful, but distinguish mystery from confusion.
- Reveal character through decisions under pressure, not labels alone.
- Use scene-level turns that alter goals, relationships, knowledge, or power.
- Do not make every chapter end with a gimmick; an emotional reframe can be a re-hook.
Essays, articles, and educational prose
- Use a problem, misconception, contradiction, case, or question as the central loop.
- Give readers necessary definitions before asking them to evaluate evidence.
- Turn through evidence or reasoning, not withheld basic context.
- Make each section change the reader’s model, not merely add another example.
- Favor clarity and scanability when readers need reference value more than suspense.
Journalism, research, legal, medical, and financial writing
- Lead with the verified finding or decision-relevant conclusion when delay could mislead.
- State uncertainty, limitations, counterevidence, and source quality near the claim they qualify, using the source’s epistemic strength or a weaker one.
- Use narrative structure to organize evidence, not to conceal it.
- Never manufacture urgency or imply causation from correlation.
- Treat accuracy, informed choice, and traceability as hard gates.
Marketing, sales, and case studies
- Ground stakes in real customer problems and documented outcomes.
- State the offer and audience clearly.
- Use specific proof before superlatives.
- Do not use fake scarcity, bait-and-switch hooks, unverifiable testimonials, or hidden conditions.
- Make the call to action a logical consequence of the story, not an abrupt demand.
Short-form video and social content
- Establish subject, relevance, and tension in the first spoken line or first visible beat.
- Show while telling when a real visual can confirm the claim.
- Use one dominant question and a small number of clean micro-loops.
- Cut greetings, credentials, and background unless they create immediate trust or stakes.
- Write for breath, edit points, captions, and image changes without turning every line into a fragment.
Long-form video, podcasts, speeches, and presentations
- State the destination early enough to orient the audience.
- Re-establish stakes and the active question at major section changes.
- Use stories, demonstrations, and examples to vary cognitive load.
- Place the strongest reframe near a structural turning point, not only at the end.
- End with a resolved idea, image, decision, or line the audience can repeat.
Review protocol
When reviewing existing writing, diagnose before rewriting.
1. State the reader contract
In one or two sentences, identify what the piece appears to promise, to whom, and why they should care.
2. Map attention
Mark or describe:
- hook;
- stakes;
- central question;
- causal spine;
- escalation points;
- turns;
- payoffs;
- re-hooks;
- likely drop-off points;
- ending.
Useful labels are [HOOK], [STAKES], [QUESTION], [CAUSALITY], [TURN], [PAYOFF], [REHOOK], [RHYTHM], [VOICE], and [TRUTH].
Do not insert labels into the rewritten final copy.
3. Prioritize findings
Separate findings into:
- Must fix: truth risk, broken promise, incoherent causality, missing payoff, severe confusion, or voice violation.
- High leverage: weak opening, vague stakes, predictable structure, dead transition, unearned turn, or flat ending.
- Polish: rhythm, diction, compression, imagery, and line-level emphasis.
Support each major finding with a specific excerpt or beat. Only flag truth risks present in the text or source; do not invent hypothetical misreadings. Avoid generic advice.
4. Revise at the correct scale
- If the premise is weak, fix the lens or reader promise.
- If the middle drags, fix causality, escalation, or loop spacing.
- If a reveal disappoints, improve setup or payoff rather than adding hype.
- If prose feels flat, fix specificity and rhythm after structure.
- If the voice is wrong, restore diction, syntax, humor, restraint, and point of view.
5. Return the requested deliverable
- For a review, provide diagnosis and prioritized edits.
- For a rewrite, provide the improved piece first unless the user asks for commentary first.
- For a punch-up, change only the requested sections.
- For alternatives, provide meaningfully different lenses or structures, not synonyms.
Quality rubric
Use this rubric for reviews or internal validation. Score each dimension from 0 to 5.
- Promise clarity: Is the subject and value clear early?
- Stake relevance: Does the intended audience have a reason to care?
- Question strength: Can the reader form a concrete prediction?
- Causal momentum: Do beats follow through because, but, and therefore rather than accumulation?
- Escalation: Does new material change pressure, options, or understanding?
- Turn quality: Are reframes surprising, fair, and coherent?
- Payoff quality: Are major promises resolved proportionally and clearly?
- Re-hook quality: Do transitions sustain momentum without open-loop spam?
- Tone and rhythm: Does the prose sound human, intentional, and varied?
- Ending: Does the piece stop on a complete, memorable, earned note?
Interpretation:
- 45–50: exceptional control;
- 39–44: strong and ready after light polish;
- 32–38: effective but uneven;
- 24–31: structurally weak;
- 0–23: rebuild the reader contract and causal spine.
Truthfulness is a hard gate, not a score. Any material fabrication, misleading omission, unsupported certainty, or bait-and-switch fails the piece regardless of total.
Anti-patterns
Watch for these failure modes:
- Vague bait: curiosity language without a concrete subject or question.
- Context dump: background arrives before the reader knows why it matters.
- And-then pileup: events or points accumulate without causality.
- Open-loop spam: too many unanswered questions compete for memory.
- Withheld essentials: basic information is hidden only to manufacture suspense.
- Random twist: the reveal has no fair setup or causal connection.
- Fake escalation: adjectives intensify while consequences stay unchanged.
- Perpetual tension: no orientation, relief, reflection, or payoff.
- Late re-hook: the next question opens after the reader has already reached a stopping point.
- Summary deflation: the ending repeats the whole piece after its strongest line.
- Generic voice: the rewrite could belong to any creator or brand.
- Mechanical rhythm: sentence lengths vary by formula rather than meaning.
- Visual mismatch: the image contradicts, decorates, or delays the spoken premise.
- Evidence theater: citations, numbers, or anecdotes imply support they do not provide.
- Neuroscience laundering: speculative brain language is used to make craft advice sound proven.
- Narrative overreach: a reference, policy, or instruction is made harder to use merely to feel dramatic.
Internal questions
Before finalizing, answer internally:
- What exact promise does the opening make?
- Who or what is at risk, and why does it matter to this audience?
- What is the central question?
- What will a reasonable reader predict?
- Which beat changes that prediction?
- Why does each major beat follow from the previous one?
- Which loops are open right now?
- Where is each loop paid off?
- Is the turn earned by earlier information?
- Does the next section begin before momentum dies?
- What can be removed without reducing understanding or impact?
- Did any revision distort facts, uncertainty, characterization, or voice?
- Is the final line stronger than the paragraph after it?
Transformation examples
Flat chronology → causal momentum
Before
We launched the campaign on Monday. Then traffic increased. Then the checkout failed. Then the team paused the ads. Then sales recovered.
After
We launched on Monday, and traffic doubled by noon. But the extra load exposed a checkout failure, so every new visitor increased the number of abandoned carts. We paused the ads, fixed the bottleneck, and relaunched that evening. Traffic returned. This time, sales did too.
Vague hook → specific question
Before
You will not believe what happened when we changed one thing.
After
We cut the signup form from nine fields to three. Conversions fell 18 percent. The question was not whether shorter forms work—it was what those six deleted fields had been doing for trust.
Random surprise → earned turn
Before
The knight reached the dragon, but suddenly the dragon was friendly.
After
The knight raised his sword. The dragon backed away, dragging something through the ash. Only then did he see the chain around its neck. It had never been guarding the gate. It had been chained to it.
Dead transition → re-hook
Before
The prototype passed the test. Next, we looked at manufacturing.
After
The prototype passed every test—which made the manufacturing estimate impossible to ignore.
Hype → truthful stakes
Before
This mistake will destroy your business if you do not fix it today.
After
If refunds are rising while support tickets stay flat, customers may be leaving before they ask for help. That gap can hide a product problem until the next revenue report.
Monotonous rhythm → intentional rhythm
Before
The room was quiet. The screen was blank. The team was waiting. The deadline was close. The build was failing.
After
The room was quiet. On the blank screen, the build timer kept moving while nobody spoke. Two minutes remained. The build failed again.
Output contract
When this skill is active:
- Deliver the requested prose, script, outline, rewrite, or review—not a lecture about storytelling.
- Keep internal beat maps, predictions, and scoring private unless the user asks to see them.
- Do not mention this skill in the finished creative work.
- Preserve formatting constraints, citations, and source boundaries supplied by the user.
- Explain only the highest-leverage changes unless a detailed critique is requested.
- When multiple approaches are valid, choose a strong default and provide alternatives only when they add real value.
- Stop when the reader contract is fulfilled. Do not add a weaker paragraph after the strongest ending.
What ships with it: 29 files
3160.1 KB alongside SKILL.md, 10 of them executable
comparison/
- corpus.json64.7 KB
- README.md3.0 KB
- render-table.mjsruns4.3 KB
- results.json245.0 KB
- run-pairs.mjsruns20.4 KB
- TABLE.md220.3 KB
review-app/
- app/csv-utils.mjsruns375 B
- app/globals.css22.0 KB
- app/layout.tsx1.2 KB
- app/page.tsx119 B
- app/ReviewWorkspace.tsx30.2 KB
- eslint.config.mjsruns465 B
- .gitignore505 B
- next.config.tsruns133 B
- .openai/hosting.json91 B
- package.json1.2 KB
- postcss.config.mjsruns94 B
- public/og.png2271.3 KB
- public/results.json245.0 KB
- README.md913 B
- scripts/sync-results.mjsruns3.1 KB
- tests/review-workspace.test.mjsruns3.2 KB
- tsconfig.json666 B
- vite.config.tsruns1.7 KB
- worker/index.tsruns1.7 KB
- CHANGELOG.md2.1 KB
- .gitignore58 B
- LICENSE11.1 KB
- README.md5.1 KB