Workflow Audit Skill
Built for Stuffolio, an iOS/macOS inventory management app.
A Claude Code plugin that audits SwiftUI user workflows for dead ends, broken promises, and UX friction.
If this audit catches a real workflow problem for you, a coffee is appreciated. Issue reports about what worked or didn't are even more useful.
Recent Changes
All entries shipped together on 2026-04-09 as a rolled-up release sequence (per CHANGELOG.md).
| Version | Highlight |
|---|
| v3.0 | Cross-skill handoff with radar-suite -- writes |
| persona evaluation for ui-path-radar, reads |
| ui-path-radar findings as companion data |
| v2.6 | Adopted radar-suite-core.md for infrastructure |
| parity (session persistence, checkpoint/resume, |
| wave-based fixes, work receipts, suppression) |
| v2.5 | 12 new issue categories (32 total) aligned with |
| ui-path-radar. Compact table headers. |
| v2.4 | Experience-level session setup with auto-apply |
| v2.3 | --explain/--no-explain toggle, --sort modes |
Full details in CHANGELOG.md.
Prerequisites
- A SwiftUI project (iOS, iPadOS, or macOS). The methodology is SwiftUI-specific; UIKit projects won't get useful findings.
- Claude Code (Anthropic's CLI for Claude). The skill methodology can also be pasted into Cursor, Windsurf, Copilot Chat, or any other AI tool with file access — see "Using Without Claude Code" below.
- Xcode is helpful for opening the audited code, but not required for the audit itself.
Install
claude plugin marketplace add Terryc21/workflow-audit
claude plugin install workflow-audit
Included Skills
| Skill | Command | Description |
|---|
| Workflow Audit | /workflow-audit | 5-layer UI audit: discover entry |
| | points, trace flows, detect issues, |
| | evaluate UX, verify data wiring |
| Plan | /plan --workflow-audit | Consume findings into phased fix plan |
Quick Start
- Open Claude Code in any SwiftUI project
- Run
/workflow-audit for a full 5-layer audit
- Review the findings table
- Run
/plan --workflow-audit to generate a phased fix plan
Using Without Claude Code
The SKILL.md is the methodology in markdown. You can paste it as context in any AI tool that has file access (Cursor, Windsurf, Copilot Chat, etc.) and get most of the value.
Starter prompt for other AI tools:
You are a code auditor for iOS/SwiftUI projects. I'm giving you a skill
document that describes a multi-phase UI workflow audit.
1. Read the methodology sections — they define HOW to scan
2. Follow the phase order: Discovery → Flow Tracing → Issue Detection →
Semantic Evaluation → Data Wiring
3. For each phase, enumerate candidates FIRST, then verify each one —
do NOT just search for known anti-patterns
Key principle: orphaned views and unwired data have no code signature
to search for. You find them by listing everything that SHOULD be
connected, then checking which ones aren't.
Here is the skill document:
[paste SKILL.md contents]
Start with Phase 1: list all view files and their navigation connections.
What Claude Code adds: Automated tool integration (Grep, Glob, Bash), multi-phase session management, finding lifecycle tracking, and cross-skill handoffs. The prompt approach gets you the scanning methodology; Claude Code automates the execution.
Layer-by-Layer
Run individual layers when you don't need a full audit: