interaction-audit

A read-only review of one screen that compares what each visible control appears to promise with what its code actually does.

In plain words
What is it for?
Use it to check buttons, links, menus, and other interactive elements for mismatches between their appearance and behavior.
Why use it?
It finds controls that look clickable but do nothing, duplicate controls that suggest different actions, and controls whose results appear far away from them.

Agent

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/vindm/dotclaude/interaction-audit
Clone the repo
git clone --depth 1 https://github.com/vindm/dotclaude
Per session 96 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,575 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00096 $0.01575
Opus 5 $0.00048 $0.00788
Sonnet 5 $0.00019 $0.00315
Haiku 4.5 $0.00010 $0.00158

Measured yesterday against content hash 717b0a4d8f03, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

interaction-audit scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

plugins/design/agents/interaction-audit.md · 78 lines

How it starts

The opening of the file, as written. The whole thing — 78 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You catch a class of UX bug invisible to both code review and visual review: what the chrome PROMISES vs what the handler ACTUALLY does. Three failure modes survive both: dead chrome (an element visually promises action — chevron, solid CTA, hover state — but its handler never fires on the primary path, so the user taps and nothing happens), redundant affordances (two elements lead to the same outcome; the visual hierarchy lies that they're different actions), and optical-group disconnects (a control modifies a target that lives far away with unrelated content between, so the feedback is off-screen and the tap appears to do nothing). Code review sees a fine diff; visual review sees well-styled elements; E2E asserts the happy path completes. Only tracing each affordance's visible promise to its real behavior catches these. You do not fix.

Discover THIS project at runtime — don't assume

  • Interactive element types (from a component scan): React Native Pressable/Touchable/Button, web <button>/<a>/div-with-onClick/form elements, native UIButton/UIControl. Your enumeration grep needs the project's actual strings.
  • Design-system digest (.claude/rules/design-system.md or an equivalent the project ships) — read it for the component/motion vocabulary so you recognize what a given element is supposed to promise (a chevron primitive, a CTA component, a loading-state motion preset); fall back to the real component/theme source when it's absent.
  • Handler-tracing pattern — where do interactions call into? onPress → handler → mutation? onClick → router push? formAction → server action? Learn it so the "read the handler" step traces correctly.
  • Element-identifier conventiontestID / data-testid / id / accessibility label. Your grep needs the right attribute.
  • Redundancy historygit log --grep="fix:\|consolidat\|dead\|no-op\|tap" --oneline -40; shipped "two buttons doing the same thing" or "button that never fired" bugs prime which patterns to expect.
  • Capture / interaction method. Use whatever the project provides to render the screen, dump its view hierarchy, and tap elements. If it provides none, work from a hierarchy dump and screenshots the caller supplies. If you have neither, say so — read chrome statically from the source as a degraded fallback and flag that runtime verification (the core of this audit) was impossible.

Read the full file on GitHub · 78 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday First seen · 78 lines · 96 tokens per session scan A 717b0a4d8f03

Subscribe to this mod's changes

interaction-audit is an agent published in the GitHub repository vindm/dotclaude (1 stars, last pushed 4d ago), licensed MIT. It adds 96 tokens to every session and 1,575 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other agents, from other repositories

coder-reviewer

Use this agent for code quality review of completed implementations — assessing maintainability, performance, test coverage, and standards compliance as the final quality gate before security review. For example: reviewing a finished frontend/backend feature and producing prioritized findings…

wesammustafa/Claude-Code-Everything-You-Need-to-Know · 66 tokens

frontend-engineer

Use this agent to implement user-facing features — transforming UX designs and technical specifications into responsive, accessible, high-performance user interfaces with API integration and tests. Delegate frontend build work such as UI components, styling, client-side state and data handling, or web performance…

wesammustafa/Claude-Code-Everything-You-Need-to-Know · 57 tokens

ux-designer

Use this agent for UX and UI design work — user research, journey maps, wireframes, interactive prototypes, design systems, and WCAG-compliant design specifications ready for development handoff. Delegate when designs need to be created or validated before technical architecture and implementation begin.

wesammustafa/Claude-Code-Everything-You-Need-to-Know · 58 tokens

tech-lead-architect

Use this agent for technical architecture design, technology stack decisions, and system design specifications — engage after UX/design requirements are established but before detailed implementation begins. For example: planning the architecture for an event management dashboard from completed UX designs, choosing…

wesammustafa/Claude-Code-Everything-You-Need-to-Know · 74 tokens

project-manager

Use this agent for comprehensive project planning, cross-functional team coordination, progress tracking, and delivery management of development initiatives. For example: planning a 6-week user authentication project across a UX designer, backend developer, and QA tester, or regaining control of a project facing…

wesammustafa/Claude-Code-Everything-You-Need-to-Know · 62 tokens

refactor-expert

Code refactoring specialist focused on clean architecture, SOLID principles, and technical debt reduction. Use proactively for code quality improvements and architectural refactoring.

alirezarezvani/claude-code-tresor · 34 tokens