discover-apple-check

name	discover-apple-check
description	Use for iOS projects to pre-validate against Apple App Store Review Guidelines before development. Fills the rejection-risk checklist in start-project.md.

Discover — Apple Check

Overview

Pre-development Apple App Store review. Instead of discovering rejection risks after 3 months of coding, this sub-skill surfaces them before the first commit. Invokes the apple-appstore-reviewer skill, cross-references lessons from the swift-calories wing (the one app that passed Apple review), and consolidates all known Apple rejection patterns — including the newer ones Apple announces piecemeal in separate review cycles.

Standalone or part of /vladyslav:discover. Fills section 11 of docs/product/start-project.md.

Scope: iOS only. Do not run for web / Android / non-Apple projects.

Type: Architect

Process

Step 1: Verify iOS project

Run <plugin>/scripts/detect-stack.sh . and parse JSON. If .ios is false → STOP: "This skill is iOS-only. Detected stacks: ." (List from the JSON keys that are true.)

If .ios is true → record the project name from CLAUDE.md and proceed to Step 2.

Step 2: Read context

Read:

• docs/product/start-project.md — sections 1 (ідея), 2 (проблема), 3 (audience), 4 (MVP scope), 7 (tech constraints)
• CLAUDE.md — project rules
• Any existing docs/product/apple-review.md

Step 3: MemPalace — pull swift-calories lessons

Crucial: swift-calories is the wing of the one app that actually passed Apple review. Its lessons are the canonical reference for "what Apple actually rejects."

Run these searches as one parallel batch (independent reads — see _shared/references/orchestration-conventions.md) and surface ALL hits to the user:

mempalace_search wing=swift-calories "apple review"
mempalace_search wing=swift-calories "rejection"
mempalace_search wing=swift-calories "guideline"
mempalace_search wing=swift-calories "privacy manifest"
mempalace_search wing=swift-calories "IAP"
mempalace_search wing=swift-calories "demo account"

Also cross-wing:

mempalace_search "apple rejection"
mempalace_search "app store review"

Compile the findings into a "known-risk list" to feed Step 4.

If swift-calories has zero apple-review records, warn the user:

"The swift-calories wing has no apple-review records seeded yet. This sub-skill will still run, but the output will only contain generic Apple guideline advice — not the curated rejection patterns from the one app that actually passed review.

Options:

1. Pause here, run /vladyslav:ingest in the swift-calories project to populate the wing, come back — recommended
2. Continue with generic advice only (weaker, but functional)
3. Abort"

Proceed based on their choice. Do not silently skip this — the warning is load-bearing.

Step 4: Invoke apple-appstore-reviewer

The apple-appstore-reviewer skill lives at ~/.claude/skills/apple-appstore-reviewer/SKILL.md. Read it via the Read tool (not Skill tool — it's a plain prompt-based skill, not a plugin skill) and apply its guidance to the idea + MVP scope described in start-project.md.

Feed the reviewer this input:

You are auditing a planned iOS app BEFORE development begins. Review against Apple App Store Review Guidelines:

Idea: <section 1 of start-project.md>
Problem: <section 2>
Audience: <section 3>
MVP scope: <section 4>
Tech constraints: <section 7>

Known-risk list from swift-calories wing (the one app that passed review):
<insert MemPalace findings from Step 3>

Produce a prioritized list of rejection risks, covering ALL of these areas (not just the obvious ones):

1. Guideline 4.2 — minimum functionality (is this "web wrapper" or "simple list" territory?)
2. Guideline 5.1.1 + 5.1.2 — privacy: what data, privacy manifest, third-party SDK declarations
3. Guideline 3.1.1 — payments: IAP vs external for digital goods, no workarounds
4. Guideline 2.1 — demo account: will reviewers need credentials? (auth / paywall / premium features)
5. Guideline 1.2 — UGC moderation: reporting, blocking, takedown
6. AI-generated content disclosure (2024-2025 rejection pattern)
7. Guideline 5.1.5 — location access, background modes, justification
8. Guideline 4.0 — design: accessibility, dark mode, dynamic type, iPad support claims
9. Guideline 2.3 — metadata: screenshots must match, no placeholder content

For EACH risk, produce:
- Severity: high / medium / low
- What to decide NOW (before coding) — architectural decisions that lock in compliance
- What to verify DURING development — checkpoints in add-feature
- What to verify BEFORE submission — in pre-release-check

CRITICAL: Apple gives feedback piecemeal across separate review cycles. Consolidate ALL known patterns upfront so the first submission covers everything at once.

Step 5: Synthesize the rejection-risk checklist

Produce a structured section 11 replacement with:

• Risk status: <green | yellow | red> based on severity count:
  • GREEN: 0 high-severity risks, ≤2 medium
  • YELLOW: 1 high-severity OR 3+ medium
  • RED: 2+ high-severity
• High-severity risks: bullet list with "decide now" actions
• Medium-severity risks: bullet list
• Low-severity risks: bullet list
• Demo account plan: if auth is in MVP, plan credentials now
• Privacy manifest plan: list the data types and third-party SDKs that will need declaration
• Link to full analysis: docs/product/apple-review.md

Step 6: Update files

1. Overwrite section 11 of docs/product/start-project.md with the risk summary. Preserve all other sections.
2. Create/overwrite docs/product/apple-review.md with the full apple-appstore-reviewer output plus the consolidated risk list.

Step 7: MemPalace record

mempalace_check_duplicate first. If new, mempalace_add_drawer:

• wing: current project (NOT swift-calories — each project gets its own record)
• room: decision
• content:

  [WHAT] Pre-development Apple review for <project>
  [STATUS] <GREEN|YELLOW|RED>
  [HIGH RISKS] <count> — <one-line per risk>
  [DECISIONS LOCKED IN] <bullets of architectural decisions>
  [FILES] docs/product/apple-review.md, docs/product/start-project.md#11
  [SOURCE] swift-calories wing + apple-appstore-reviewer
  [DATE] <today>
  

• added_by: discover-apple-check

Also: if you discovered NEW rejection patterns not already in the swift-calories wing, write them there too so the knowledge base grows:

• wing: swift-calories
• room: problem
• content: [WHAT] New Apple rejection pattern: <name>. [DETAILS] <what triggers it>. [DISCOVERED DURING] <project> pre-review, <date>.
• added_by: discover-apple-check

Step 8: Architect report

✓ Architect report — Discover Apple Check
- Risk status: <GREEN|YELLOW|RED>
- High-severity risks: <count>
- Medium: <count>
- Low: <count>
- Decisions locked in: <count>
- Demo account planned: <yes|no>
- Privacy manifest items: <count>
- Files updated:
  - docs/product/start-project.md (section 11)
  - docs/product/apple-review.md (new)
- MemPalace records: <project> wing + <count> new patterns in swift-calories wing

Next steps:
- /vladyslav:add-feature — start building once Apple risk status is GREEN or YELLOW (high-severity resolved)
- /vladyslav:design-sync — canonize the design system before committing to screen layouts

Notes

• The swift-calories wing is the canonical Apple knowledge base. Treat it as the source of truth for "what Apple actually rejects in 2025+". When you discover new patterns, write them back there.
• Apple gives feedback piecemeal. A first submission might get rejected for reason A, a second for reason B. This sub-skill's job is to front-load ALL known reasons so the first submission covers everything — turning the review cycle from N-iterations into 1.
• AI-content disclosure is a new pattern (2024-2025). Apps that generate responses/content via LLMs must disclose this in the UI, not just in the app description. Rejection is silent if missed.
• Don't run this on non-iOS projects. The Step 1 check is a hard stop — Apple guidelines don't apply to web or Android.