Creative logic reconstruction
Skill zacfire/creative-logic-reconstruction-skill/creative-logic-reconstruction
Use when the user asks why a product, work, creator, team, company, or strategy is shaped the way it is, what logic drove it, or what can be learned from it. Covers product logic, creative intent, presentation logic, creator/team context, interviews, tradeoffs, and transferable lessons.From its SKILL.md
npx -y skills add zacfire/creative-logic-reconstruction-skill --skill creative-logic-reconstructionAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 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.
SKILL.md
9.6 KB, ~2.1k tokens by cl100k_base, as published. Nobody here has run it
Creative Logic Reconstruction
Core Idea
This skill reconstructs the logic behind a product or work from its visible form plus creator/team context. It is an outside-view reconstruction, not an internal retrospective.
It is not a summary, review, or praise exercise. The goal is to answer:
Why did they make it this way?
What scene, audience, constraint, and belief made this form rational?
What did they optimize, sacrifice, and learn?
What can be transferred, and what should not be copied?
Use For
- Product, app, agent, tool, website, service, or feature logic analysis.
- Films, shows, videos, books, essays, talks, newsletters, games, art, or creator content.
- Team/founder/creator strategy interpretation from public or provided evidence.
- “分析这个团队/创作者为什么这么做”, “还原这个作品的呈现逻辑”, “它有什么可学的”.
Evidence Discipline
Separate every major claim into:
| Status | Meaning |
|---|---|
observed | Directly visible in the product/work/source material. |
inferred | Reasonable reconstruction from observed evidence and context. |
unverified | Claimed by marketing, media, or memory but not verified. |
refuted | Evidence contradicts the claim. |
Do not turn preference into intent. Do not say “they wanted X” when the evidence only supports “this design optimizes for X”.
Research Order
- Primary object first: inspect the product/work itself, screenshots, artifact, text, code, structure, narrative, UI, pacing, affordances, constraints.
- Creator/team context second: official site, docs, release notes, changelog, interviews, founder posts, talks, podcasts, funding/team background, previous works.
- Audience and market third: user comments, reviews, community response, competitors, category norms.
- Synthesis last: reconstruct goals, tradeoffs, operating logic, and transferable lessons.
For current companies, teams, interviews, release status, pricing, or public claims, browse or use available primary/local sources. If exact source material is not available, state the uncertainty.
Reconstruction Framework
0. Mode Selection
First choose the analysis mode. Do not force every subject into a product template.
| Mode | Use When | Emphasize |
|---|---|---|
| Product / App | software, services, agents, tools, hardware | user scene, entry point, loop, trust, distribution, business constraints |
| Work / Content | film, book, essay, video, art, game, speech | structure, pacing, medium choices, audience emotion, narrative payoff |
| Creator / Team | founder, author, director, studio, lab, collective | repeated choices, operating beliefs, constraints, taste, evolution |
| Strategy / Investment | company, category, market, product bet | positioning, wedge, moat, timing, risk, validation path |
If the subject crosses modes, state the primary mode and the secondary mode.
1. Surface Form
Describe what exists before explaining why:
- Product/work shape.
- Entry point.
- First impression / first scene / first use.
- Core loop.
- Output or payoff.
- What is unusually emphasized or hidden.
2. Target Scene
Infer the concrete scene it is optimized for:
Who is using/reading/watching this?
When?
With what urgency, mood, knowledge, or constraint?
What are they trying to avoid?
3. Creator Goal Hypothesis
State likely goals as hypotheses, not facts:
Likely goal:
Evidence:
Why this form helps:
What it sacrifices:
Confidence:
For major inferred motives, include evidence and confidence. Prefer:
Hypothesis:
Evidence:
Alternative explanations:
Confidence:
4. Core Tradeoffs
Look for decisions where they chose one value over another:
- Speed vs depth.
- Clarity vs richness.
- Control vs magic.
- Trust vs autonomy.
- Mass market vs expert users.
- Narrative emotion vs factual completeness.
- Originality vs genre convention.
- Execution breadth vs quality.
5. Operating Logic
Explain how the product/work moves the user/audience through time:
For products:
entry -> context -> decision/action -> feedback/status -> memory/result -> next use
For writing/books:
promise -> framing -> evidence -> tension -> reframing -> conclusion -> aftertaste
For film/video:
hook -> world/character setup -> conflict -> rhythm shifts -> emotional turn -> payoff
5a. Misclassification Check
Before applying lessons, ask:
What does this appear to be?
What is it actually optimizing?
What wrong category would mislead us?
Examples:
A screenshot inbox may look like a photo manager but actually optimize post-capture action.
A desktop agent may look like chat but actually optimize execution orchestration.
A sensor accessory may look like content capture but actually optimize behavior signals.
A slow film may look uneventful but actually optimize attention, unease, or moral ambiguity.
Use this check whenever the surface form may hide the real product or creative logic.
6. Hidden Constraints
Infer constraints that likely shaped the form:
- Technical constraints.
- Distribution/channel constraints.
- Team size and skill mix.
- Trust/safety/regulatory pressure.
- Cost/latency/model limits.
- User education burden.
- Commercial positioning.
- Brand or aesthetic identity.
7. What They Believe
Name the underlying belief only when evidence supports it:
They seem to believe that [mechanism] matters more than [alternative].
Example:
They seem to believe desktop context is more valuable when explicitly summoned than continuously observed.
8. Transferability
Always separate:
| Category | Meaning |
|---|---|
| Learn directly | Mechanism can be adopted with similar goals. |
| Translate | Useful, but must be adapted to our product philosophy. |
| Do not copy | Works for them because their positioning differs. |
| Watch later | Interesting but not relevant to current phase. |
9. Contextual Transfer
When the user has an active project, product, work, company, thesis, or decision context, translate lessons into that context. Do not assume the context is always the user's own product or any specific project.
If no active context is named, give general lessons only.
Separate:
| Category | Meaning |
|---|---|
| Directly transferable | Mechanism can be reused with little adaptation. |
| Needs translation | Useful but must be adapted to the current context. |
| Do not copy | Attractive surface choice that depends on different positioning. |
| Open question | What must be verified before applying it. |
10. Behavioral Reconstruction Guardrails
Use evidence-based reconstruction methods, not psychological profiling.
Use:
- Repeated choices over isolated details.
- Competing hypotheses before choosing one explanation.
- Opportunity / constraint / capability analysis.
- Anomaly analysis: what is unusually emphasized, omitted, avoided, or repeated.
- Confidence calibration for every major inferred motive.
Do not:
- Diagnose personality, pathology, or private traits.
- Infer sensitive traits, mental health, criminality, or private motives about real people.
- Infer hidden intent from sparse evidence.
- Treat aesthetic preference as proof of motive.
- Turn creator/team background into gossip.
- Say “they wanted X” when evidence only supports “this form optimizes for X”.
Output Template
Use this unless the user asks for a different format. Adapt headings to the selected mode; do not force every heading onto films, books, essays, art, or creator analysis.
## One-Line Thesis
## What It Is Actually Optimizing
## Surface Logic
## Creator / Team Goal Hypotheses
## Core User / Audience Scene
## Key Design or Creative Tradeoffs
## Operating Logic
## Evidence Map
Mark important claims as `observed`, `inferred`, `unverified`, or `refuted` in the Evidence Map or inline.
## Risks and Blind Spots
## Misclassification Check
## Contextual Transfer
## What We Should Not Copy
For long-form research docs, add:
## Source Notes
## Open Questions
## Follow-Up Research
Common Mistakes
-
Mistake: Summarizing what it says instead of reconstructing why it is shaped that way. Fix: Ask what scene, constraint, or belief makes this form rational.
-
Mistake: Treating marketing claims as true intent. Fix: Compare claims against product/work behavior.
-
Mistake: Overfitting to one clever feature. Fix: Reconstruct the whole operating loop.
-
Mistake: Copying the mechanism without copying the positioning. Fix: Identify what must be translated for the user’s product/philosophy.
-
Mistake: Assuming the relevant context is always the user's current product. Fix: Only map lessons to a specific context when the user names it or the conversation clearly establishes it.
-
Mistake: Using “profiling” language to sound insightful. Fix: Reconstruct behavior from evidence, list alternatives, and calibrate confidence.
-
Mistake: Using team background as gossip. Fix: Only use background when it explains visible choices, constraints, taste, distribution, or strategic posture.
Style
Be direct and calibrated. Use “likely”, “seems”, and “the evidence supports” when reconstructing intent. Challenge weak inferences. Prefer diagrams and tables when they clarify the logic.
What ships with it: 1 file
360 B alongside SKILL.md
agents/
- openai.yaml360 B