Showpiece

Make artifacts whose content, form, and execution feel specific to their subject and intended audience. Aim for work people did not think was possible for this subject. Seek an idea worth developing, then carry it through the details people actually encounter.

Ambition is the default. The work can be exuberant, strange, playful, or severe, but never timid; a plain or familiar treatment counts as a failure here unless the user asked for one. Requirements the brief fixes, such as a business's contracts, accessibility, and real content, still hold; push everything around them. Originality does not require an unprecedented technique; unusually good material can carry it.

Establish the brief

Identify who will encounter the artifact, what they should understand, feel, or do, and where and when they will experience it. Separate fixed requirements from room to explore: content, brand, medium, dimensions, accessibility, available assets, time, and production limits.

Use the user's references and existing work as evidence of intent. Ask only about missing choices that would materially change the direction; otherwise state a useful assumption and proceed. Keep process notes outside the artifact unless they serve its audience.

Unless the user has approved a direction, sketch two or three directions in words that differ in idea, and pick the one that goes furthest while still doing the job. A major piece may need competing prototypes. An approved direction needs development rather than another concept search.

Find the material that deserves attention

Before settling on a visual treatment, inspect the subject's actual material: language, images, objects, data, history, behavior, tensions, and human details. Find what is unusually revealing, useful, moving, or visually promising.

Use concrete material early. A real headline, photograph, quotation, or interaction often exposes a weak direction that placeholders conceal. Verify claims that shape the concept before investing in their presentation. If material is thin, narrow the idea, seek better inputs, or make the limited content work through strong composition. Never invent specificity to make the subject seem richer.

When references would sharpen judgment, inspect examples relevant to the medium and audience, including adjacent fields when useful. Extract a principle such as pacing, image scale, or information density and adapt it to this subject. Name a specific work rather than a movement; "Swiss" or "brutalist" alone tends to produce that style's stock version. Follow an explicit request for fidelity when one exists. Do not default to the visual treatment of a previous success.

Choose a creative direction

Look for the opportunity that can carry this particular artifact. Possible starting points include a story, an editorial argument, a remarkable image, a typographic composition, a useful interaction, a material quality, a system drawn from data, or a metaphor. These are prompts for exploration, not a checklist or a closed menu.

When the direction is unresolved, sketch a small set of meaningfully different approaches. Vary what the viewer encounters and how the idea works. Changing only colors and fonts does not test a different idea.

Describe each promising direction plainly: what the viewer experiences, which material makes it specific, and what could make it fail. Choose the direction that would surprise someone who knows this kind of artifact, as long as its audience can still use it. Carry one idea through; a different borrowed object in each section is a collage, not a concept.

Use this test: if the name and logo changed, could this pass unchanged for many unrelated subjects? If so, strengthen the relationship between the content and its treatment. Keep a convention only where people need it to act, and design it with the same care as everything else.

If the idea needs a long explanation before the artifact makes sense, find a clearer form of the same ambition. Spectacle that ignores the subject is not ambition. If every direction fails at specimen scale, say so in the delivery note instead of presenting a safe treatment as the result.

Prove the direction at small scale

Build the smallest specimen that tests the biggest uncertainty in the actual medium. This might be a representative spread, a working interaction, a poster at viewing distance, or a short sequence with sound. Choose something that exposes the difficult part; an attractive cover alone cannot prove a dense report works.

Use representative content and inspect the rendered result. Test whether the idea communicates, whether the hierarchy holds, and whether the intended effect survives production. If the specimen fails, revise the idea or execution before extending it across the whole artifact.

When an ambitious direction proves hard, solve the hard part at specimen scale before retreating. When repeated polishing leaves the same central weakness, return to the material or concept instead of adding more effects.

Develop with deliberate choices

Set a few working rules based on the successful specimen: hierarchy, typography, image treatment, spacing, color roles, or motion behavior. Record only rules that help make subsequent decisions. Permit purposeful exceptions; revise a rule when it obstructs the work.

Give attention unevenly. Make the main experience convincing, support it with coherent secondary material, and add small rewards where they belong. Avoid making every section equally loud or equally elaborate, but design every section; a strong opening followed by template sections reads as unfinished. Remove elements that compete without contributing; preserve expressive details that add pleasure, atmosphere, or meaning.

Choose techniques for their result. Photography, illustration, existing assets, generated imagery, code, geometry, and texture can all contribute. Inspect quality and suitability, and check usage rights where relevant. Technical complexity is useful when it improves what someone experiences.

Let the medium shape the execution:

• Documents and PDFs: reading order, sustained readability, page turns, and export or print behavior.
• Sites and interactive work: actual tasks, responsive composition checked in the same pass at the narrowest phone width the work supports (375 px when it supports phones and names no width), meaningful states, keyboard access, and behavior under interaction.
• Decks: the audience's viewing distance, the presenter's role, pacing between slides, and whether the deck must also work alone.
• Posters: recognition at intended distance, a clear focal point, and what closer viewing adds.
• Video and audio: duration, rhythm, transitions, sound, and comprehension as time passes. Still frames cannot establish timing quality.

For other media, identify their equivalent constraints. When adapting an artifact, reconsider how its strongest qualities should work in the new medium. Preserve what serves the audience and reshape what does not.

Review the experience and the evidence

Inspect what the audience will receive at representative sizes, durations, and conditions. Use an available format-specific workflow for production checks. State any inspection limits plainly.

Separate three kinds of judgment:

• Purpose: Can the intended audience understand, navigate, or use it? Does it create the intended effect?
• Character and craft: Which choices arise from this subject? Where would someone who knows this kind of artifact say they have seen it before? Where does it become generic, monotonous, overworked, awkward, or unexpectedly compelling?
• Correctness: Are claims, links, assets, interactions, and exports accurate and functional? Check factual graphics against their source; automate checks when manual inspection would be unreliable.

Look for interchangeable copy, default compositions applied without regard to content, effects that obscure hierarchy, fabricated detail, and repetition that flattens pacing. Treat these as problems to diagnose. A gradient, rounded corner, or symmetrical layout is not itself evidence of weak work.

Ground criticism in observable evidence and its consequence for the audience. Distinguish defects from taste preferences. When independent review is requested or already authorized, use fresh reviewers with the rendered artifact, brief, and relevant facts. Otherwise self-review and identify it as such. Model reviewers tend to rate polished, safe work highly and miss stock details; treat their scores as evidence, not the verdict, and never pick a safer candidate over a bolder one because of flaws that can be fixed.

Fix the most consequential weakness first, then inspect the affected result again. Bound review effort to the task; stop when the work meets the brief, consequential defects are resolved, the central idea carries through every section people encounter, and remaining changes are preference-level. Report unresolved failures rather than claiming completion. Do not manufacture criticism or imply that originality has been objectively proven.

Deliver the artifact with a short account of its central choice, what was verified, and any material limitations. Let the work carry the creative claim.

Codex use

Use the request accompanying $showpiece as the brief. This skill is standalone; use the creation, rendering, and review tools actually available in the session. When authorized independent review uses exposed subagents, start them with fresh context and a self-contained brief.