ios-engineering

Reusable open-source AI skill for the full lifecycle of an existing SwiftUI/iOS product: planning, implementation, review, debugging, UI validation, focused XCTest generation, whole-app audits, and evidence-based verification.

This repository is built for agent runtimes that load a SKILL.md entry point together with reusable references, templates, and helper scripts. The skill stays product-agnostic on purpose: project-specific module names, build schemes, domain rules, and validation commands belong in the target repository, not in the shared skill.

Why this skill

• Use one reusable skill surface for product planning, implementation, review, debugging, testing, auditing, and release verification.
• Keep project facts separate from reusable engineering guidance, so the skill can move across teams and codebases cleanly.
• Route work into focused modes instead of mixing planning, debugging, review, and verification into one vague prompt.
• Preserve evidence, output structure, and validation expectations for higher-signal agent runs.

Preview

Command surface

Command	Use it when
/ios plan <requirement>	Product structure, UX, data loop, execution grouping, or acceptance criteria are not frozen yet.
/ios implement <task>	You need a production change within a defined scope.
/ios review <scope>	You want a defect-focused engineering review of completed or partial work.
/ios debug <failure>	The observed behavior is wrong and the root cause is still unknown.
`/ios ui <replicate	review
/ios test feature <feature>	You want focused XCTest coverage for one feature or contract.
/ios audit app	You want staged whole-app QA or release-candidate validation.
/ios verify <requirement>	You need a final pass, conditional pass, or fail verdict with evidence.

Detailed workflows live in modes/. Reusable engineering guidance lives in references/. Project-specific facts should stay in the target repository's profile, requirements, domain rules, design system, and testing instructions.

Installation

Keep the repository intact after installation so relative links to modes/, references/, templates/, and scripts/ continue to work.

Codex runtime

git clone https://github.com/greysonOuyang/ios-engineering-skill.git \
  "${CODEX_HOME:-$HOME/.codex}/skills/ios-engineering"

Claude-style runtime

git clone https://github.com/greysonOuyang/ios-engineering-skill.git \
  "$HOME/.claude/skills/ios-engineering"

If your setup uses another runtime folder such as ~/.agents/skills, swap the destination path and keep the final directory name as ios-engineering.

Source checkout plus symlink

git clone https://github.com/greysonOuyang/ios-engineering-skill.git \
  "$HOME/Projects/ios-engineering-skill"

ln -sfn "$HOME/Projects/ios-engineering-skill" \
  "${CODEX_HOME:-$HOME/.codex}/skills/ios-engineering"

Update an installed copy

cd "${CODEX_HOME:-$HOME/.codex}/skills/ios-engineering"
git pull --ff-only

Restart your agent runtime if it caches available skills.

Example prompts

/ios plan "Add iPad-ready recurring schedule editing without changing existing creation behavior"
/ios implement ISS-142 --focus=data
/ios review current branch
/ios debug "After editing a recurring item, today's timeline becomes empty"
/ios ui review SettingsScreen
/ios test feature ReminderComposer
/ios audit app --scope=release-candidate
/ios verify "Recurring plan editing is complete"

Natural-language equivalents work too. If the user explicitly names a mode, the skill should route into that mode first.

Repository layout

ios-engineering/
├── SKILL.md
├── COMMANDS.md
├── modes/
├── references/
├── templates/
├── scripts/
├── agents/openai.yaml
├── metadata.json
└── assets/

• SKILL.md: entry point, routing, and operating rules
• COMMANDS.md: compact command reference for mode invocations
• modes/: focused workflows such as plan, implement, review, debug, test, audit, and verify
• references/: reusable engineering guidance loaded only when a task needs it
• templates/: durable report, prompt, and verification scaffolding
• scripts/: helper tooling for repeatable analysis
• agents/openai.yaml: UI metadata for OpenAI-compatible skill surfaces
• metadata.json: package metadata for distribution, indexing, and cataloging

Project profile expectations

Start from templates/project-profile.md when a target repository does not already expose:

• product and domain truth locations
• module ownership
• state and persistence owners
• design-system names and component paths
• schemes, targets, and validation commands
• project-specific runtime invariants

Compatibility and boundaries

• Designed for existing native iOS or SwiftUI applications, not greenfield web stacks.
• Expects a runtime that supports skill directories with a SKILL.md entry point.
• Keeps web-specific view-layer workflows, such as Tailwind or React refactors, outside this skill.
• Avoids hard-coding product names, token names, build targets, or domain models.

Contributing

Issues and pull requests are welcome when they improve reusable iOS engineering workflows, references, templates, or helper scripts without baking in project-specific behavior.

License

MIT