interface-critique

Files (5)

• LICENSE.txt
• NOTICE.md
• SKILL.md
• openai.yaml · agents/
• heuristics.md · references/

SKILL.md (first lines)

# Interface Critique

Resolve one stable target and its primary user goal. Read relevant product context and inspect both source and rendered behavior when available.

This skill owns holistic design diagnosis and prioritization. It is not an exhaustive file-and-line standards audit and does not implement fixes unless requested.

## Independent assessments

Run two assessments without allowing either to anchor the other:

1. **Design assessment:** specificity, hierarchy, composition, information architecture, typography, color, emotional fit, discoverability, states, content, and cognitive load.
2. **Evidence assessment:** keyboard and touch behavior, semantics, accessibility tree, responsive layouts, console/runtime signals, and deterministic source checks.

If isolated sub-agents are available and authorized, run both in parallel. Otherwise complete them sequentially and label the report `single-context`.

Use [references/heuristics.md](references/heuristics.md) for the shared evaluation baseline. Project standards override generic preferences.

## Synthesis

Combine evidence only after both assessments finish. Identify agreement, disagreement, false positives, and missing runtime proof.

Report:

1. Method and target.
2. Design-specificity verdict.
3. Overall impression and strongest opportunity.
4. Two or three strengths.
5. Three to five priority issues.
6. Relevant persona failure paths.
7. Minor observations.
8. Questions that could unlock a stronger direction.

For each issue include severity (`P0`–`P3`), exact evidence, affected user, why it matters, and a concrete fix. Be direct; do not inflate preferences into violations.

Do not modify the interface unless the user also asks for fixes.

From deepaktiwari09/team-agent-skills, MIT.

More from deepaktiwari09/team-agent-skills