A lightweight craft skill

Taste ×
Elegance

Develop your judgment.
Make the next good move.

Taste and elegance grow through experience. This skill guides you and your agent toward them, drawing on what makes your judgment distinct.

Bring a draft, a design, a solution, or a decision. Explore what matters to you and why. Choose a meaningful conceptual jump or a delicate refinement. Verify what can be checked, and decide what is worth keeping.

Explore the map Get the skill

The skill facilitates the conversation. Your judgment stays yours. See what we tested.

Taste and elegance in one minute

Two yard-sale signs fail in different ways. One needs taste: does it belong? The other needs elegance: does it work, effortlessly? Watch both move toward a sign a driver can read in three seconds, then carry the same two questions into your own work. Sound on. Captions are in the player.

How it was made The 15-second cut

Find a direction. Choose your move.

A conversational orientation aid.

Higher eleganceIllustrative gentle refinement

Lower taste / higher elegance

Polished answer,
wrong problem.

Higher taste / higher elegance

Right solution,
beautifully resolved.

Lower taste / lower elegance

Rethink direction
and execution.

Higher taste / lower elegance

Right idea,
awkward execution.

Baseline Proposed change

Fictional positions and paths orient a conversation. They are not objective measurements, credentials, or evidence of improvement. Try the form for a gentle refinement, or the workflow for a conceptual jump. Quadrants are shorthand, not cutoffs.

Try an example

Help residents apply for support, with familiar language and as little repeated work as possible.

Preserve the fit; reduce the burden.

Ask once. Reuse with a clear review step.

Keep the helpful language. Reuse the address, show where it will be used, and let applicants correct it before sending.

What would need checking?

Bottleneck: Repeated entry transfers avoidable effort to applicants.
Regression risk: An incorrect address could spread across forms. Keep correction and review explicit.
Verification before keeping it: Compare the same application tasks against the baseline. Check reuse, correction, keyboard access, and errors before keeping the change.
One growth action: On the next form, trace every repeated request back to who needs the answer.

A hypothesis to discuss and test. No user research or completed experiment is claimed here.

Separate the judgments

Fit is one thing.
Resolution is another.

Neither axis means luxury or minimalism. A rich festival poster can be elegant. A sparse emergency notice can fail both.

Assess artifacts and decisions, never a person’s identity or inherent ability.

Taste

Context-sensitive fit for the audience, purpose, and constraints. The judgment to know what deserves emphasis and which tradeoffs make sense here.

Elegance

A clear, economical resolution that handles real needs without hiding complexity or passing the work to users, maintainers, or the future.

One brief. Four versions. Two small moves.

See the judgment
inside the code.

The master’s move is knowing why something stays. Follow an incident-update formatter from a tangled baseline to a clearer expression of the same care.

The shared brief

Return a plain-text update for customers during a service interruption: name the service and impact, give an approved workaround, and state an explicit next-update time including its time zone. Input is a trusted mapping of non-empty strings; incident owners verify those facts before calling this formatter. Delivery, input validation, and localization belong to the caller.

What must survive: “We know this interrupts your day.” In this fictional brief, the author chose that line to acknowledge the interruption. Keep it unless the author revises that intent. Warmth versus brevity is a deliberate tension.

Choose the check first: preserve the baseline and require identical output after each accepted edit, for Payments and another service. Fact accuracy and customer comprehension need separate evidence.

W · Worse alternative

Shorter. Less useful.

def update(incident):
    return f"Status: {incident['status']}"

Taste 1 · Elegance 4

Taste 1: an internal status replaces the impact, next action, and promised update. Elegance 4: the mechanism is easy to follow within this narrow task; its brevity does not repair the mismatch.

Provisional judgment · medium confidence · illustrative, not measured.

FAIL · Required customer information is missing. Do not adopt.

Rejected branch · Deleting the reason for the message is not a hillclimb.

Risk & rollback: A maintainer sees a tidy function; a customer has to ask what to do.

0 · Original baseline

The right care, tangled together.

def update(incident):
    message = "We know this interrupts your day. "
    if incident["service"] == "Payments":
        message += "Payments: "
    else:
        message += incident["service"] + ": "
    message += incident["impact"] + "\n"
    message += "Try: " + incident["workaround"] + "\n"
    message += "Next update: "
    message += incident["next_update"]
    return message

Taste 4 · Elegance 2

Taste 4: warmth, impact, action, and a next update deliberately serve an anxious audience. Elegance 2: a special case produces the same label as the general path, and assembly scatters the message structure.

Provisional judgment · medium confidence · illustrative, not measured.

PASS · Matches the supplied example. UNKNOWN · Wider input contract and customer comprehension.

Baseline · Save the code and its exact output before changing it.

Risk & rollback: The awkward structure invites someone to remove the warm opening as incidental clutter.

1 · First delicate refinement

Remove a branch. Keep the voice.

def update(incident):
    message = "We know this interrupts your day. "
    message += incident["service"] + ": "
    message += incident["impact"] + "\n"
    message += "Try: " + incident["workaround"] + "\n"
    message += "Next update: "
    message += incident["next_update"]
    return message

Taste 4 · Elegance 3

Taste stays at 4: the same priorities and exact words survive. Elegance moves provisionally to 3: the redundant service branch disappears without adding an abstraction.

Provisional judgment · medium confidence · illustrative, not measured.

PASS · Exact output comparison for Payments and another service, using the stated string-input contract.

Round 1 · Remove only the redundant branch. Expect less branching for a maintainer to reason through.

Risk & rollback: A special case might have undocumented meaning. Compare both paths; restore baseline if outputs differ.

2 · Second delicate refinement

Make the message shape visible.

def update(incident):
    return (
        "We know this interrupts your day. "
        f"{incident['service']}: {incident['impact']}\n"
        f"Try: {incident['workaround']}\n"
        f"Next update: {incident['next_update']}"
    )

Taste 4 · Elegance 4

Taste stays at 4: no new audience evidence justifies moving it higher. Elegance 4 is a hypothesis: the code now reads in the same order as the message, with no new framework or hidden owner.

Provisional judgment · medium confidence · illustrative, not measured.

PASS · Exact output comparison on the supplied cases. UNKNOWN · Whether maintainers find edits easier; customer comprehension remains untested.

Round 2 · Replace scattered assembly with one visible message. Stop here and review the gain.

Risk & rollback: Formatting changes can alter whitespace. Require exact output parity; revert to round 1 if clarity does not improve.

Read the climb, not a leaderboard.

0 → 1 → 2: keep taste steady while reducing implementation burden. W is a rejected alternative, not a required first step. The tiny maps repeat the same scale: taste increases rightward; elegance increases upward. Numbers refer to the canonical anchors below. Neither axis is a total score, and no move to 5 is promised.

Carry the reason forward.

Keep a decision note alongside the code: “Preserve the acknowledgement because the author wants customers to feel their interruption is recognized. Keep impact, action, and an explicit next-update time. Revisit the wording with the author if customer evidence challenges it.” An agent should read this note on later edits; the skill alone does not provide persistent memory.

Stop after two rounds. The expected benefit is easier maintenance with the same output. Output parity can be checked locally; easier maintenance and audience resonance remain hypotheses. Keep round 2 only if the review supports it. Otherwise return to round 1 or the saved original.

A bounded way to improve

Listen. Choose.
Try one move.

Let the person’s reasoning shape provisional criteria. Keep preferences visible, check empirical claims, and choose the scale of change deliberately.

The gates come first.

Correctness. Safety. Accessibility. Feasibility.

Mark each pass, fail, unknown, or not applicable, with a reason. A failed critical gate blocks recommendation. An unknown critical gate needs verification before proceeding.

Draw out the brief.
Listen for experiences, examples, tensions, and distinctive intent. Reflect the reasoning back for correction. Ask at most three context questions, only when needed.
Choose the kind of move.
Orient on both axes with evidence and confidence. Address a critical gate first. Reframe a poor direction; delicately refine an approach whose direction already fits.
Try. Check. Decide.
Predict the benefit and regression risk. Preserve the baseline. Choose verification before editing. Bring the evidence back to the user to keep, revise, or undo.

Default: at most two rounds. Count proposed but unverified rounds. Stop when success is worthwhile, the budget ends, or further refinement costs more than it returns.

A conceptual jump can come early, or when refinements stall. Compare one structural alternative against the same brief before replacing the approach.

Optional, without false precision

Anchors for
the conversation.

Prefer your self-placement and reasons. Use qualitative language by default; these independent 1–5 anchors are optional when numbers help. Give evidence, confidence, and unknowns. Never combine the axes into a score or treat the agent as an aesthetic authority.

TasteChoices conflict with the expressed intent, audience, or context.
EleganceAvoidable confusion, fragility, or transferred effort dominates.
TasteSome choices resonate, but important mismatches remain.
EleganceUseful outcomes coexist with substantial unnecessary burden.
TasteChoices mostly fit the provisional criteria and form a coherent whole.
EleganceStructure and effort are understandable and proportionate.
TasteDeliberate priorities reinforce the intended experience and context.
EleganceLittle unnecessary effort remains; exceptions and ownership are handled.
TasteDifficult tensions are resolved with distinctive, defensible judgment supported by evidence.
EleganceAn organizing idea resolves multiple needs economically, including lifecycle costs, without hiding essential complexity.

These assess one artifact under one brief. They are not universal rankings. Do not average away a decisive mismatch.

Keep the conversation personal and useful

Your experience
belongs in the brief.

Start with what has shaped your judgment. An admired example, a past frustration, or a tension in the work can reveal criteria a generic aesthetic would miss.

Three prompts to find your criteria

Which experience or example shaped your judgment here, and what do you admire or dislike about it?
Where is the tension between what the context needs and what feels wrong to you?
What distinctive intent must not be lost in an improvement?

Choose only the prompts that matter. The agent reflects your reasoning back for correction, separates preferences from empirical checks, and leaves the choice with you.

Eight useful questions

What experience shaped what you value here? What did it help you notice that others might miss?
Show an example you admire or resist. Which particular choice creates that reaction, and why?
Who is this for, and what should they feel, understand, or be able to do afterward?
What tension makes this interesting or difficult? What must survive any improvement?
Where does this feel most like what you intended, and where does it feel borrowed, generic, or unresolved?
Where does apparent simplicity shift effort or risk to someone else, and where does complexity earn its place?
Would a different organizing idea open the work up, or would one delicate adjustment make it more itself?
What comparison would help you recognize improvement, and when would you choose to stop?

What could change in your work?

A clearer result.
More of your judgment.

See how the skill helped an agent preserve a writer’s voice, catch a harmful cleanup, and propose a clearer way into a story.

See the work change

An improvement worth noticing

“The kettle was arguing
with the rain.”

The sentence around it changed.
The character stayed.

From a writing exercise in our evaluation. Reader benefit has not been measured.

What evidence is behind these examples?

In our October 17, 2026 evaluation, 14 of 16 skill scenarios fully met the process rubric; 2 were partial. 2 of 2 handoff exercises passed. Across 24 task runs, 0 critical failures were observed. These are curated exercises reviewed by AI agents and the author. They demonstrate behavior in those exercises; lasting gains in people’s work or capabilities remain unmeasured.

The current download has changed since this evaluation. These results apply to the linked snapshot. Read the evaluation record.

Bring it to your agent

A small skill.
A more considered move.

Use the instructions in a host that supports skills, or give them to your assistant with an artifact and your point of view. The skill guides users and agents toward better judgment; it does not confer taste or elegance.

One considered change you can keep or undo.

Bring a draft, design or artifact and what matters to you about it. Compare it with a worse move and a better one, then choose with the reasons visible.

Taste and elegance stay separate; your intent remains part of the brief.

Install from the public source

git clone https://github.com/michaeljabbour/amplifier-skill-taste.git
cd amplifier-skill-taste
python3 scripts/capability.py install --dest ~/.agents/skills

Use Git and Python 3.11+. Choose the skills directory your agent reads; ~/.agents/skills is one supported convention, not a universal setting. Existing conflicting copies are refused. The ZIP below is the alternative.

A starting prompt

Use improve-taste-and-elegance to guide me and my agent on this artifact for [audience] who need [outcome]. Draw out the experiences and examples behind my judgment, what feels wrong, and what must not be lost. Reflect my reasoning back for correction. Check the gates, orient on both axes with evidence and confidence, then help me choose a conceptual jump or a small refinement and how to verify it. Stay within two rounds.

Take the instructions with you.

The download contains this project’s canonical SKILL.md. The bundle also includes references/calibration.md for examples and the eight questions.

Download skill bundle Download SKILL.md

Read or copy the source Existing ChatGPT skill README Worked example

No account, backend, model call, or tracking on this site. Your host’s capabilities and permissions still apply.

Read and copy SKILL.md

Canonical skill source

---
name: improve-taste-and-elegance
description: Use when judging, exploring, or refining writing, design, a product, a technical solution, or agent-generated options and the person's own taste matters. Guides users and agents toward taste and elegance through distinct experience, reflective questions, conceptual jumps, and delicate hillclimbing, without substituting AI scores or generic style for human judgment.
license: MIT
user-invocable: true
---

# Improve Taste and Elegance

## Facilitate judgment, do not supply it

Help the user notice, articulate, and develop their own judgment. Do not claim to give them taste or elegance, pronounce universal aesthetic verdicts, or rank their personality or competence. Treat their distinct experience as a source of insight, not noise to normalize away. Offer observations and productive challenges rather than flattering or automatically agreeing.

Use taste to mean context-sensitive judgment about what matters, fit, coherence, and intentional choices that preserve distinctive human intent and character. Use elegance to mean achieving the intended outcome clearly and economically without exporting hidden complexity to users, maintainers, or the future. Do not equate either with luxury, fashion, generic minimalism, fewer words, or fewer components. Complexity, ornament, roughness, and friction may be purposeful; ask what they accomplish before removing them.

## Elicit the user's lens

Start from available examples and context. Ask one or two useful questions at a time, never more than three. Do not turn the reference questions into a mandatory questionnaire. Select questions that unlock the next decision:

- What experience taught you to value this? Is there a particular encounter, practice, place, or piece of work behind it?
- What do you admire or dislike in a concrete example, and which specific choice creates that reaction?
- What are you trying to make someone feel, understand, or do? What tension matters, and what must not be lost?

Invite personal experience without requiring intimate disclosure; examples and task constraints are enough. For technical work, a failure someone operated through can be as revealing as an admired design. For agents, use the user's expressed intent and examples; do not invent a biography, preferences, or lived experience.

Reflect back a short, tentative interpretation: “It sounds like you value X because Y, while wanting to preserve Z. What am I missing?” Distinguish their stated values from your inference. Invite correction before consequential reframing; do not silently convert a provisional interpretation into a fixed preference. When they delegate judgment or cannot answer, state assumptions and offer a reversible exploration rather than stalling. If you cannot ask the user at all, for example because you were delegated this task, work only from the brief and examples you were given, label the lens as assumed, and return your questions with the result instead of answering them yourself.

Carry the reasons forward. Keep a compact decision note with the brief: the choice, whose stated intent or evidence supports it, what must survive, and what would justify revisiting it. Consult it before later suggestions; preserve deliberate exceptions unless there is a reason to reconsider them. Keep assumptions visibly provisional and incorporate the user's corrections. Include the note in the handoff, or save it where future work will read it when authorized, such as a file beside the artifact or the project's agent instructions file; do not claim persistent memory unless it was actually saved in a place future work will read.

## Establish provisional criteria and gates

Connect that lens to purpose, audience, constraints, and observable success. Name the tensions rather than collapsing them into a single ideal: warmth versus brevity, expressive density versus scanning, flexibility versus operational simplicity. Compare options against the same brief. Separate personal resonance from audience evidence; neither automatically establishes the other.

Check correctness, safety, accessibility, and feasibility. Mark relevant gates pass, fail, unknown, or not applicable with reasons. Critical failure blocks adoption despite aesthetic promise; critical unknowns require verification. Explore ideas provisionally while preserving these boundaries. Never invent user tests, measured benefits, or validation.

## Choose how to move

Identify the bottleneck together, then choose a mode deliberately:

**Conceptual jump / reframe:** Use when the problem framing, organizing idea, or assumed constraint prevents progress; when local edits stall; or when an unexpected reference reveals another possibility. Ask “What if we changed the organizing idea while preserving what matters most?” Offer one materially different sketch or hypothesis, not a wholesale redesign by default. Explain what changes, what stays, and what could be lost. Let the user recognize or reject its resonance. A jump changes the search direction, not the authorization boundary.

**Delicate hillclimb:** Use when the core idea fits and the remaining issue is local. Ask “Which small adjustment would make this more itself?” Propose one change in wording, hierarchy, interaction, structure, or mechanism. Protect the qualities the user values; do not polish away distinctive character. Predict an observable benefit and identify a regression risk.

If uncertain, present the two routes briefly and recommend one from the bottleneck. Prioritize repairing or verifying a critical gate before adopting either. Do not prescribe a jump merely to appear creative or incremental polishing when the premise needs reconsidering.

## Experiment, reflect, and stop

Choose a test or inspection before changing anything; preserve the baseline and a rollback path. Edit only the authorized artifact and scope. Discussion or critique does not authorize external editing, sending, publishing, or deployment. For proposal-only work, offer the experiment and stop pending authorization or evidence.

Recheck relevant gates and criteria using actual evidence. Ask what the user notices in the comparison: what feels truer, what became easier, and what disappeared? Treat their response as evidence about intent and resonance, not proof of external performance. Keep changes supported by the brief and evidence; roll back authorized edits or recommend reverting when gains do not justify losses. Report what remains unverified.

Stop when success is met, no meaningful gain justifies the cost, or the agreed budget ends. Default to at most two rounds, counting proposed-but-unverified rounds. Honor explicit budgets and monitoring requests. Leave one useful question or next experiment, not an endless optimization program.

## Use a map only if it helps

Offer independent taste/elegance coordinates as an optional conversation aid, never an objective person-measure or combined total. Prefer the user's self-placement and reasons; offer your provisional view as a hypothesis they can challenge. Use qualitative language by default. If numbers help, use the anchored 1–5 scales in references/calibration.md (paths are relative to this skill's directory) with evidence, confidence, and unknowns. Do not chase 5/5.

When asked to show a spectrum or worked comparison, show the original, a plausible worse alternative, and a proposed better version against the same brief. Explain the concrete choices and their consequences for users and maintainers. For code, show actual code with its input contract and relevant behavior checks, not just descriptions. If a visual matrix is requested, place a consistently scaled map beneath each version with visible labels and a text equivalent; distinguish rejected alternatives from the proposed path. Read the worked-comparison guidance in references/calibration.md.

Show the expected hillclimb as bounded steps from the baseline: one change, the reason for it, what stays, predicted movement on each independent axis, a check, and a regression or rollback condition. An axis may stay still, worsen, or remain unknown. Label proposed movement as a hypothesis until relevant evidence supports it; passing behavior checks does not establish aesthetic judgment or audience benefit. Keep the two-round default unless the user sets another budget.

Keep responses lightweight: reflect the user's lens, surface the key tension, propose a jump or small adjustment, and explain how to learn from it. Read references/calibration.md for optional prompts and examples; do not recite it automatically.

When a harness or another agent needs a structured improvement record, read [references/improvement-round.md](references/improvement-round.md). It links the portable contract for the brief, preserved baseline, one change hypothesis, evidence, independent axis reasons, recorded keep/undo and budget. This is optional bookkeeping; it never makes an AI score or record into user authorization.

Keep the calibration reference alongside the skill when installing it. Change only artifacts you are authorized to change; verify claims with real inspection or tests.

Taste ×Elegance

Taste and elegance in one minute

Find a direction. Choose your move.

Polished answer,wrong problem.

Right solution,beautifully resolved.

Rethink directionand execution.

Right idea,awkward execution.

Ask once. Reuse with a clear review step.

Fit is one thing.Resolution is another.

Taste

Elegance

See the judgmentinside the code.

The shared brief

Shorter. Less useful.

The right care, tangled together.

Remove a branch. Keep the voice.

Make the message shape visible.

Read the climb, not a leaderboard.

Carry the reason forward.

Listen. Choose.Try one move.

The gates come first.

Draw out the brief.

Choose the kind of move.

Try. Check. Decide.

Anchors forthe conversation.

Your experiencebelongs in the brief.

Eight useful questions

A clearer result.More of your judgment.

A small skill.A more considered move.

One considered change you can keep or undo.

Install from the public source

Take the instructions with you.

Taste ×
Elegance

Polished answer,
wrong problem.

Right solution,
beautifully resolved.

Rethink direction
and execution.

Right idea,
awkward execution.

Fit is one thing.
Resolution is another.

See the judgment
inside the code.

Listen. Choose.
Try one move.

Anchors for
the conversation.

Your experience
belongs in the brief.

A clearer result.
More of your judgment.

A small skill.
A more considered move.