AGENT SKILL

ad-story-loops

Plan the question-and-payoff structure of an ad, promo, product demo or social marketing video so viewers keep watching for a reason and leave with the message: reveals written first, one central question, earned surprise, consequence, closure. Use when a brief asks for retention, hooks, open loops, story loops, curiosity gaps, rehooks, cliffhangers or a series of episodes, or when a cut loses viewers. Pairs with commercial-beat-sheet, explainer-storyboard, motion-ad, ugc-video and script-video, which own the beats and the look.

License 许可
AGPL-3.0
最近更新
2026-11-15
Stars
★ 635
Forks
91

01 前言 Front Matter

name
ad-story-loops
description
Plan the question-and-payoff structure of an ad, promo, product demo or social marketing video so viewers keep watching for a reason and leave with the message: reveals written first, one central question, earned surprise, consequence, closure. Use when a brief asks for retention, hooks, open loops, story loops, curiosity gaps, rehooks, cliffhangers or a series of episodes, or when a cut loses viewers. Pairs with commercial-beat-sheet, explainer-storyboard, motion-ad, ugc-video and script-video, which own the beats and the look.

03 SKILL.md 正文

Story loops for ads and marketing video

A story loop is one question the viewer wants answered and the payoff that answers it. This skill produces the loop map: which questions the piece opens, where each one pays off, and what each payoff changes. Write it before the beat sheet and store it on the board. It decides the order of information, not the look and not the runtime split.

The board skill still owns the beats and timecodes: commercial-beat-sheet for a spot, explainer-storyboard for an explainer, motion-ad for a sound-off feed ad, ugc-video for a creator ad, script-video when the voiceover sets the timing. Load the one that fits, plan the loops here, then fill its beats in the loop map's order.

The rule

Every question the ad opens earns a meaningful answer, and every answer changes what the viewer understands about the product or the problem. A reveal that changes nothing is noise. A question that never pays off is bait, and the viewer learns not to trust the next one.

What the evidence supports

Users bring creator frameworks (Kallaway's Story Loops and Dopamine Ladder, Paterson's TAVIN, open loops, the Pixar Story Spine). Use them as writing heuristics. None is a validated formula, and "dopamine" is not a measurement anyone can make on an ad. What controlled research does support:

  • Curiosity about a specific, answerable question makes people wait for the answer and remember it better (Loewenstein 1994, Kang et al. 2009, Gruber et al. 2014).
  • Cliffhanger endings raise the stated wish to see the next installment, but not enjoyment or suspense (Media Psychology, 2023). Unfinished stories are not more memorable: a 2025 meta-analysis found no general recall advantage for interrupted tasks.
  • Knowing an outcome can raise enjoyment (Leavitt and Christenfeld 2011). Viewers value how and why something happens, not only what happens, so a piece does not need constant surprise.
  • Absorption in a character's situation (narrative transportation) moves attitudes. Curiosity alone does not carry an ad.

So: frame claims as questions where it helps, vary the kind and timing of payoffs, keep the causal chain readable, and close the central question. Never write "dopamine" or "addictive" into a brief, a report or ad copy as if it were a result.

Write the reveals first

  1. List the true reveals. One to three facts from the brief or the product page that would make a viewer say "oh": a number, a before and after, a mechanism, a trade-off solved. Each must be a confirmed fact. A reveal the brief cannot support is [CLIENT INPUT NEEDED], never invented.
  2. Name the expectation each reveal overturns. A surprise only lands when the viewer already holds the expectation it breaks. "The robot lifts 200 kg" means little. Show it straining at a small box first, and the jump becomes the payoff. The fact stays the same. The setup gives it meaning.
  3. Work backward to the setup. For each reveal, write the minimum the viewer must see or know first: the old way, the cost, the failed attempt, the competitor default. That setup is what the early beats show.
  4. Pick the central question. One question the whole piece answers, in the viewer's words: "Can one photo really become a usable ad?" Everything else is a nested loop under it.

The seven stages

Each loop moves through stages 2 to 6. A piece runs stage 1 once, repeats 2 to 6 per loop, and ends on 7. This sequence is a synthesis of the frameworks above, not a validated model.

Stage Job On screen
1. Relevance Who or what matters, and why to this viewer The person, the problem or the stake, mid-action
2. Prediction A clear, answerable question The setup the reveal will overturn. A line that names the doubt
3. Development Evidence, a choice, or a complication The attempt, the mechanism, the obstacle
4. Payoff The answer, or a real advance on it The reveal, shown rather than claimed
5. Consequence What changed and why it matters The result in use. The trade-off now solved
6. Next question Open a new loop only when another is warranted A new doubt the payoff created
7. Satisfaction Enough closure to earn trust The central question answered, then the CTA

Consequence is the stage that separates an ad from a trick: the twist has to change the viewer's picture of the product. Satisfaction is the stage the endless-loop model drops. Stop reopening questions and finish.

Loop budget

Starting points for feed and pre-roll ads, not research results. Adjust to the platform and the product.

Runtime Loops First payoff by
6–15s The central question only ~55% of runtime
30s Central question plus one nested loop ~45%
60s Central question plus two nested loops ~40%
90s and longer Central question plus up to four, each closed before the next opens ~30%

A payoff held too long frustrates rather than holds. Never have more than two loops open at once, and never open a loop in the last 15% of a piece unless it is a series hook (below).

Mapping onto the board skills

Fill the board skill's beats in the loop map's order. Typical fits:

Board skill Relevance + question Development Payoff + consequence Satisfaction
commercial-beat-sheet Hook, brand landing Problem Solution/USP, proof CTA, reiteration, closing echo of the hook
explainer-storyboard Problem, stakes Shift Mechanism beats, the aha, outcome/proof CTA, sign-off
motion-ad Hook headline over the product in context Before state, the tap Before becomes after, proof line End card
ugc-video Hook to camera The thing happening physically The result on the creator Verdict to camera

commercial-beat-sheet's "close every loop you open" and its hook archetypes still apply. Choose the archetype that sets up the strongest reveal: Result First and Cost Reveal open a "how?" loop, In Media Res Failure and Cold Contradiction open a "what went wrong?" loop.

Hooks that open a loop, and hooks that only bait

A loop-opening hook names a specific doubt the viewer can hold: "We tried to make a full product ad from one photo without editing a frame." The viewer knows the challenge and can predict an outcome.

Bait withholds the subject itself: "You won't believe what happened", a blurred product, a "wait for it" with no stake. It raises a question the viewer has no reason to care about, and the payoff cannot change their understanding of anything.

A mid-piece rehook ("it looked professional, until we noticed one detail") works only when the detail pays off within a few seconds and matters to the buyer. Do not script a reversal the product's real story does not contain.

People, not only questions

Product facts alone rarely move anyone. Give the viewer someone whose goal they understand, the creator, the customer, the team, and show what the product changes for them. Keep two things apart when writing:

  • Character commitment: how much the person on screen wants the outcome.
  • Audience investment: whether the viewer cares. It comes from recognising the problem as theirs and seeing what is at stake.

A large on-screen sacrifice does nothing if the viewer does not know why it matters. Put the stake in stage 1, in the viewer's terms.

Series and episodic content

For a run of marketing videos (a launch week, a build-in-public series, a how-we-made-it set), each episode closes its own central question, then opens the next episode's question in the last beat. Cliffhangers raise intent to watch the next one. They do not make this one better, and an ad that ends unresolved leaves the buyer without the answer the CTA depends on. Use an open ending only when a next episode exists and is scheduled, and still pay off the episode's own question first.

Across episodes, repeat what the viewer can rely on: the same presenter, the same format, answers that arrive. Loyalty to a creator or brand comes from consistent value, expertise and trust, not from escalating suspense.

Store the loop map on the board

The loop map lives on the storyboard so a later pass, or a variant, can read it:

  • Put the whole map in the board brief with create_storyboard or edit_storyboard set_board {brief}, under a LOOP MAP heading.
  • Tag each shot's loop job in its notes field (add_shot or update_shot): L1 open: can one photo make a usable ad?, L1 payoff, L1 consequence, central close. notes never enters a render prompt, so it costs nothing in the picture.
LOOP MAP
CENTRAL: <the question the whole piece answers, in the viewer's words>
REVEALS: <1–3 confirmed facts, each with its source>
L1  expectation: <what the viewer assumes>
    question:    <the doubt the shot names>
    opens:       <timecode, shot slug>
    payoff:      <timecode, shot slug, what is shown>
    consequence: <what the viewer now understands differently>
L2 ...
CLOSE: <timecode, shot slug, how the central question is answered>
OPEN AT END: <none, or the next episode's question>

The shot text the renderer reads (action, motion) still follows the chosen model's prompting skill. The loop map changes what each shot shows and in which order, not how the prompt is worded.

Review the cut against the map

After assembly, check each loop on the real frames: preview_timeline_frame {timeline_id, times_ms, sheet: true} at every question and payoff timecode in the map, then view_image on the sheet.

  • Is the expectation visible before the reveal? A reveal with no setup on screen is a non-sequitur.
  • Does each payoff arrive by its planned timecode, and is it shown rather than claimed in a super?
  • Is the consequence on screen, the product in use after the reveal?
  • Does the question read muted? On a sound-off feed ad the question is a headline (caption-titles), not only voiceover.
  • Is the central question answered before the CTA, and is nothing left open that the brief did not ask to leave open?

Name each failure against its loop (L2 payoff has no setup), fix the board or the cut, and preview the same timecodes again.

Worked example: a 45-second product demo

The central question: can one product photo become a usable, editable ad?

Time Stage Shot
0–5s Relevance, L1 question "We tried making a full product ad from one photo, without editing a frame." The photo on screen
5–13s Development, L2 opened Generation starts. "It looked surprisingly professional, until we noticed one detail."
13–22s L2 payoff The generated ad has changed the product's logo. It looks good and is commercially unusable
22–34s Consequence, L3 question "So we kept the product untouched and generated only the motion graphics around it." The product stays fixed while the graphics move
34–45s L1 and L3 payoff, satisfaction The finished ad beside its editable timeline. The question is answered, the trade-off shown, then the CTA

Each reveal changes the viewer's understanding of the problem: the obvious method fails for a reason a buyer cares about, and the fix is the product's real mechanism.

When the piece is done

ad-creative-testing designs the matched variants that show whether a loop structure actually beats a direct version for this product and audience. Retention alone is not the result. A cut people finish but do not understand or act on has entertained the wrong audience.

04 同仓库更多 Skills

nodetool-ai
nodetool-ai/nodetool

humanizer

Rewrite AI-sounding text so it reads like the writer without changing what it says. Use when editing or reviewing prose for AI tells: not-X-but-Y contrasts, one-line closers, staged openers, forced triads, dashes everywhere, inflated claims, sales language, stock AI words, bold labels, or filler.

★ 635
nodetool-ai
nodetool-ai/nodetool

nodetool

Use the NodeTool MCP server to build, validate, run, and debug visual AI workflows, and to work with assets, collections, and jobs. Use when the user mentions NodeTool or asks for a workflow graph, a node, or a media generation run on their NodeTool server.

开发工程师Skill与智能体开发
★ 635
nodetool-ai
nodetool-ai/nodetool

wan-2-6-prompting

Prompt Alibaba's Wan 2.6 video — text-to-video with a global style line plus timing-bracketed shots, and image-to-video that describes only what changes over time. Covers the shot-mode switch per provider, the prompt budget, negative prompts, motion intensity wording, audio input and prompt expansion. Use whenever the model id names a Wan 2.6 video route (alibaba/wan-2.6/text-to-video, alibaba/wan-2.6/image-to-video, wan-video/wan-2.6-t2v, wan-video/wan-2.6-i2v, wan-video/wan2.6-i2v-flash) on generate_video, animate_image or a TextToVideo node. Not for Wan 2.5 or 2.7, or the Wan 2.6 image routes.

图像与多媒体
★ 635
nodetool-ai
nodetool-ai/nodetool

video-workflow

Build a reusable NodeTool video workflow for batches or repeated inputs. For a single finished piece, use a storyboard skill.

图像与多媒体Skill与智能体开发
★ 635
nodetool-ai
nodetool-ai/nodetool

video-clone

Analyze a reference video and recreate its shot structure with a new product or cast in NodeTool.

图像与多媒体
★ 635
nodetool-ai
nodetool-ai/nodetool

video-audio-continuity

Keep sound continuous across a multi-scene piece cut from generated video — why one clip per scene hard-cuts the audio at every boundary, when to write all the scenes into a single generation instead, the sound-design vocabulary that survives a provider's copyright filter, and how to check the length, the aspect ratio and the mix of what came back. Use when a piece has more than one scene and the video model writes its own audio (Seedance 2, Veo 3, MiniMax H3, Kling 2.6 and later), or when a cut's sound drops out at the shot changes.

图像与多媒体
★ 635