Imported from garethpaul/HomeScreen (AGENTS.md). Install upstream with npx skills add garethpaul/HomeScreen. Copyright stays with the author.

AGENTS.md

Repository purpose

garethpaul/HomeScreen is a Apple platform application or Objective-C/Swift sample. HomeScreen app for for sharing your #Homescreen

Project structure

• Makefile - repository verification targets
• scripts - baseline checks and helper scripts
• docs - plans, notes, and generated README assets
• HomeScreen.xcodeproj - Xcode project
• Crashlytics.framework - repository source or sample assets
• Fabric.framework - repository source or sample assets
• HomeScreen - repository source or sample assets
• HomeScreenTests - repository source or sample assets
• TwitterKit.framework - repository source or sample assets

Development commands

• Install dependencies: no repository-specific install command is documented.
• Full baseline: make check
• Make gates support absolute checkout paths containing spaces; preserve the single-Makefile authority boundary and recursive regression.
• Local Apple development: open HomeScreen.xcodeproj
• If a command above skips because a platform toolchain is missing, verify on a machine with that SDK before claiming platform behavior is tested.

Coding conventions

• Language mix noted in the README: C/C++ headers (23), Swift (19).
• Preserve legacy Xcode project settings and signing assumptions unless the change is explicitly about modernization.

Testing guidance

• Test-related files detected: HomeScreen/UpdateStatus.swift, HomeScreenTests/HomeScreenTests.swift
• Start with the narrowest relevant test or Make target, then run make check before handing off if the change is not documentation-only.
• Keep README verification notes in sync when commands, fixtures, or supported toolchains change.

PR / change guidance

• Keep diffs focused on the requested repository and avoid unrelated modernization or formatting churn.
• Preserve public APIs, sample behavior, file formats, and documented environment variables unless the task explicitly changes them.
• Update tests, README notes, or docs/plans when behavior, security posture, or validation commands change.
• Call out skipped platform validation, legacy toolchain assumptions, and any risky files touched in the final summary.

Safety and gotchas

• Detected references to Twitter. Keep API keys, OAuth credentials, tokens, and account-specific values in local configuration only.
• The Fabric upload build phase reads FABRIC_API_KEY and FABRIC_BUILD_SECRET locally and skips the upload when either value is unset.
• Do not commit real Fabric, Twitter, signing, screenshot, or local xcconfig values to this repository.
• Home screen screenshots can reveal private apps, messages, accounts, or location hints. Keep uploads user-initiated and avoid raw response or image logging.
• Keep Twitter read-path failures category-only; do not log raw request, transport, or localized error objects.
• Keep share-composer dismissal behind validated status creation on the main queue.
• Share post callbacks are generation-bound so duplicate taps and stale completions cannot dismiss the composer.
• Keep tweet search, guest-login, model-load, and UI completion callbacks weakly owned and bound to the latest request generation.
• Treat Twitter session state as optional on presentation paths; expired or missing sessions should not crash profile-image rendering.
• This looks like an Apple platform project or sample. Xcode, Swift, CocoaPods, and deployment target versions may need to match the original project era.

Agent workflow

1. Inspect the README, Makefile, manifests, and the files directly related to the request.
2. Make the smallest source or docs change that satisfies the task; avoid generated, vendored, or local-environment files unless required.
3. Run the narrowest useful validation first, then make check or the documented package/platform gate when available.
4. If a required SDK, service credential, or external runtime is unavailable, record the skipped command and why.
5. Summarize changed files, commands run, and remaining risks or follow-up validation.