docs: add named test bundle instructions to CLAUDE.md

[kodjima33]

Summary

• Added clear instructions to both root and desktop CLAUDE.md that when testing a feature/bug, always create a separate named bundle with OMI_APP_NAME="feature-name" ./run.sh
• This prevents overwriting "Omi Dev" and allows running multiple test builds simultaneously
• Added a full "Testing with Named Bundles" section to desktop/CLAUDE.md with rules and examples

🤖 Generated with Claude Code

[greptile-apps]

Greptile Summary

This PR adds documentation-only guidance to both CLAUDE.md and desktop/CLAUDE.md instructing Claude (and developers) to always create a named test bundle via OMI_APP_NAME="feature-name" ./run.sh when testing a feature or bug fix, rather than overwriting the default "Omi Dev" install. The documented behavior was verified against desktop/run.sh — the slugify_identifier function correctly derives com.omi.<slug> from OMI_APP_NAME, matching the examples in the docs.

• CLAUDE.md: Replaces the old single-line bundle note with an explicit Named test bundles rule and updates the post-build launch instruction to reference "the named test app" instead of "the dev app".
• desktop/CLAUDE.md: Adds a new Testing with Named Bundles section (rules + example) between "App Names & Build Artifacts" and "After Implementing Changes"; updates the latter to say ./run.sh must always be used with OMI_APP_NAME.
• Minor inconsistency: The pre-existing agent-swift connect --bundle-id com.omi.desktop-dev example in the "Verifying UI Changes" section is now at odds with the new "always use a named bundle" policy — updating it would remove potential confusion for readers following the example literally.

Confidence Score: 5/5

• Docs-only change with no runtime risk; documented behavior verified against run.sh.
• Both files contain purely documentation changes. The core claim — that OMI_APP_NAME controls the bundle ID via slugify_identifier — was confirmed accurate by reading desktop/run.sh. The only issue is a pre-existing example referencing com.omi.desktop-dev that is now slightly inconsistent with the new policy, but it does not affect any code path.
• No files require special attention.

Important Files Changed

Filename	Overview
CLAUDE.md	Clarifies desktop test bundle guidance: replaces the old single bullet about custom-named bundles with an explicit "Named test bundles" rule (always set OMI_APP_NAME) and updates "launch the dev app" to "launch the named test app".
desktop/CLAUDE.md	Adds a new "Testing with Named Bundles" section with rules and examples, and tightens the "After Implementing Changes" instruction to require OMI_APP_NAME; verified against run.sh — the documented behavior (slug → bundle ID) is accurate.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Developer wants to test a feature/fix] --> B{OMI_APP_NAME set?}
    B -- No --> C["./run.sh → overwrites 'Omi Dev'\n❌ NEVER do this for feature testing"]
    B -- Yes --> D["OMI_APP_NAME='feature-name' ./run.sh"]
    D --> E["Builds /Applications/feature-name.app\nBundle ID: com.omi.feature-name"]
    E --> F["Runs side-by-side with\n'Omi Dev' and 'Omi Beta'"]
    F --> G["agent-swift connect\n--bundle-id com.omi.feature-name"]

Loading

Comments Outside Diff (1)

1. desktop/CLAUDE.md, line 156 (link)

   Stale bundle ID in agent-swift example

   The code example in the "Verifying UI Changes" section still hard-codes com.omi.desktop-dev, but the new rules in this same file say to never use the default ./run.sh without OMI_APP_NAME — so in practice the running app will always have a named bundle ID. A reader following the example verbatim would be connecting to the wrong (or non-running) app. Consider updating the comment or the value to show the named-bundle pattern:

_{Reviews (1): Last reviewed commit: "docs: add named test bundle instructions..." | Re-trigger Greptile}