Skip to content

TheGoat395/Codex-Skills / skills/creativity

creativity

Verified

Generate original ideas or escape fixation.

references/selection-validation.md

View on GitHub
# Creative Selection And Validation

Use this reference after materially different concepts have been developed. Selection should preserve originality while removing ideas that fail the objective, constraints, or reality.

## Define the comparison before scoring

Record:

- objective and audience;
- ordinary baseline;
- fixed constraints;
- minimum viable usefulness;
- required evidence or proof;
- cost of failure and reversibility;
- dimensions on which originality matters.

Do not let the most polished description set the criteria after generation.

## Screen fatal constraints

Before ranking, remove or repair candidates that:

- contradict a fixed requirement;
- depend on fabricated evidence or unavailable capability;
- reproduce the reference rather than adapt a mechanism;
- fail the central user need;
- create an inaccessible or unusable core experience;
- require an asset, permission, budget, or timeline that cannot plausibly exist;
- lose their value when stripped of decorative novelty.

A candidate may remain as a speculative branch if the missing condition is explicit and testable.

## Compare developed candidates

Use a table when tradeoffs are material:

| Candidate | Objective fit | Structural novelty | Usefulness | Coherence | Feasibility | Evidence burden | Distinctiveness | Testability |
|---|---|---|---|---|---|---|---|---|

Use qualitative judgments by default. If weights help a consequential decision:

1. set weights before rating candidates;
2. explain why each weight follows from the objective;
3. score with evidence or observable rationale;
4. test whether a reasonable weight change alters the winner;
5. keep fatal constraints outside the weighted total.

Do not use decimal precision unsupported by measurement.

## Measure novelty against a real baseline

Novelty is relational. Compare the candidate with:

- the obvious answer to this prompt;
- established category conventions;
- supplied references;
- the user's existing work;
- the other candidate families.

State what is new: mechanism, structure, interaction, interpretation, combination, delivery, or representation. A different theme on the same mechanism is a variation, not a new direction.

## Evaluate usefulness and surprise together

Ask:

- Does the idea solve or illuminate the real problem?
- Does the surprising element cause that value, reveal it, or merely decorate it?
- Can the audience understand enough to engage?
- Does the idea become more valuable after the first encounter?
- Can it support a system, sequence, or body of work rather than one moment?

Prefer ideas where usefulness and surprise reinforce each other.

## Prototype the decisive uncertainty

Do not prototype the easiest part. Identify the assumption most likely to reverse selection:

- comprehension;
- technical feasibility;
- emotional response;
- interaction discoverability;
- production burden;
- differentiation;
- evidence availability;
- performance under a key constraint.

Build the smallest artifact that can test it. Define the observation that would support, weaken, or kill the direction before running the test.

Use real users, runtime behavior, comparative artifacts, or domain evidence when available. The model's imagined reaction is not independent validation.

## Preserve minority value

The highest aggregate score can erase a radical but valuable option. Before finalizing, identify:

- the safest strong candidate;
- the highest-upside candidate;
- the simplest viable candidate;
- the candidate with the most informative test.

Select according to the actual risk posture and objective. When uncertainty is high and testing is cheap, prefer a reversible experiment over premature consensus.

## Delivery record

For a selected concept, report:

```text
Selected direction:
Organizing mechanism:
Why it beats the baseline:
Why it beats the strongest alternative:
Fixed constraints satisfied:
Main sacrifice:
Unverified assumption:
Smallest decisive test:
Fallback or branch worth preserving:
```

For exploratory work, preserve several candidates without pretending that selection has occurred.

## Research basis

- The Geneplore framework treats creative products as repeated generation and exploration under product constraints, not one-pass ideation.
- The dual-pathway model distinguishes flexible category switching from persistent exploration within a category.
- Design-fixation findings motivate evaluating structural similarity to exemplars and separating function from surface imitation.
- Tree-of-Thoughts research supports deliberate branching, evaluation, and backtracking for problems where a search structure is useful; it is not required for routine creativity.
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 198kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 194kUpdated Claude CodeCodexVerifiedInstall