Expand manual verification across web and Apple apps

[sreejithraman]

Summary

• expand manual-verify with a shared observe-act-observe workflow
• add focused browser, iOS Simulator, and native macOS verification guidance
• keep platform mechanics behind progressive references
• define isolated intermittent attempts, assertion statuses, reproduction outcomes, consequential-action limits, and evidence requirements
• require durable persistence checks and precise macOS save-command attribution

Why

Manual verification already covered web and iOS at a high level, but platform mechanics and result criteria were implicit. This makes the cross-platform boundary explicit, adds native macOS workflows, and prevents builds, screenshots, autosave, or unreproduced intermittent defects from creating false pass claims.

User effect

Agents can select realistic workflows, operate web and Apple-platform interfaces through semantic targets, verify stored results through fresh read paths, distinguish product failures from blocked or untested assertions, and report evidence consistently across web, iOS, and macOS.

Verification

• python3 scripts/check_skills.py
• python3 /Users/sree/.codex/skills/.system/skill-creator/scripts/quick_validate.py skills/manual-verify
• git diff --check
• repeated fresh native and Gemini behavioral review rounds after substantive fixes
• final staff review: no concrete behavioral defects
• final Gemini review: clean