Installable skillMIT

Inclusive Interaction Audit

Audit an interface for multi-modal interaction support. Chains: keyboard-navigation, touch-target-design, multi-modal-input, gesture-alternatives, feedback-and-status, motion-sensitivity. Use when reviewing an existing interface for interaction accessibility.

BY wanghaishengUPDATED Oct 27, 2026
INTENDED FOR✳Claude Code↗Cursor◎ChatGPTInstall in a listed agent.

instructions.md

---
name: audit
description: "Audit an interface for multi-modal interaction support. Chains: keyboard-navigation, touch-target-design, multi-modal-input, gesture-alternatives, feedback-and-status, motion-sensitivity. Use when reviewing an existing interface for interaction accessibility."
---

# Inclusive Interaction Audit

Run a comprehensive audit of how an interface supports diverse input
methods, output channels, and physical abilities.

## Process

### Phase 1: Keyboard Pass
Using **keyboard-navigation**, attempt every task using only the keyboard.
Document: unreachable elements, focus order issues, missing focus
indicators, keyboard traps, and missing skip links.

### Phase 2: Touch and Pointer Pass
Using **touch-target-design**, measure all interactive elements.
Document: undersized targets, insufficient spacing, precision-dependent
interactions, and missing tap alternatives.

### Phase 3: Input Alternatives Check
Using **multi-modal-input** and **gesture-alternatives**, verify that
every interaction has at least two input methods. Document: single-input
dependencies, disabled paste fields, gesture-only features, and
missing alternatives for complex interactions.

### Phase 4: Feedback and Motion Check
Using **feedback-and-status** and **motion-sensitivity**, verify that
all feedback reaches multiple senses and that motion is safe.
Document: colour-only indicators, missing screen reader announcements,
dangerous motion, and prefers-reduced-motion support.

## Output

Present findings as a structured report:

1. **Summary** — overall interaction accessibility rating and critical
   blockers
2. **Input method matrix** — table showing which tasks work with which
   input methods (keyboard, pointer, touch, voice)
3. **Issue list** — each issue with severity, affected input method,
   affected users, and specific fix
4. **Prioritised fixes** — ranked by number of users unblocked

Severity definitions:
- **Critical** — a task cannot be completed with a common input method
- **Major** — a task is significantly harder with certain input methods
- **Minor** — friction exists but the task can be completed
COMPATIBILITY & REFERENCES

Platform guidance

Imported from the listed AgentSkill.sh category. Complete supplied package files and notices are retained. Required external tools and credentials are not installed. No bundled code was executed during import. Individual agent task behavior has not been tested.