GB-M0-R1 — iPhone ↔ Watch6 BLE Transport PoC

[PNHD]

Goal

Create the minimal native iOS + Wear OS scaffold needed to prove bidirectional BLE/GATT transport on a real iPhone 17e and Galaxy Watch6.

Branch

feature/gb-m0-feasibility

Source packet

agent_tasks/GB-M0-R1.md

Required implementation

• iOS: Swift/SwiftUI + CoreBluetooth
• Wear OS: Kotlin, minimal diagnostic UI
• no cloud/backend
• no Wear OS Data Layer dependency
• versioned transport envelope containing protocolVersion, sessionId, seq, type, timestamp, payload
• auditable ping/ack + reconnect logs

Software-side acceptance

• both projects build
• BLE roles explicitly documented/implemented
• iPhone → Watch and Watch → iPhone round-trip paths exist
• reconnect/state logging exists
• exact real-device test instructions exist

Hardware gate

Final status MUST remain PARTIAL — HARDWARE_REQUIRED until owner tests on physical iPhone 17e + Galaxy Watch6 per docs/M0_TEST_PLAN.md.

Stop conditions

• no polished product UI
• no automatic next milestone
• no PASS from emulator/mock payloads alone
• no merge without owner/PM review

[PNHD]

CODEX IMPLEMENTATION HANDOFF — GB-M0-R1

Session: one dedicated session for this task only
Agent: Codex
Model: 5.6 Sol
Effort: High
Speed: Standard
Role: implementation owner
Workspace: PNHD/GalaxyBridge → feature/gb-m0-feasibility
Write authority: YES, only within this task scope
Sub-agents / parallel agents: OFF
Merge authority: NONE

Read first

• README.md
• docs/CURRENT_WORK.md
• docs/DECISIONS.md
• docs/FEASIBILITY.md
• docs/M0_TEST_PLAN.md
• agent_tasks/GB-M0-R1.md

Baseline gate

Before editing, report exact:

• repository
• CWD
• branch
• HEAD SHA
• git status --short
• project/build tooling detected

If branch is not feature/gb-m0-feasibility, STOP and report.

Task

Implement the minimum real-device diagnostic scaffold for iPhone 17e ↔ Galaxy Watch6 bidirectional BLE/GATT transport.

Do not use Wear OS Data Layer as the transport.

Required architecture:

1. Native iOS app in Swift/SwiftUI using CoreBluetooth.
2. Native Wear OS app in Kotlin with the minimum BLE/GATT implementation needed for the opposite role.
3. Explicitly document which side is GATT central/client and which side is peripheral/server, and why.
4. Versioned application envelope containing at least:
   • protocolVersion
   • sessionId
   • seq
   • type
   • timestamp
   • payload
5. Deterministic PING → ACK round trip in both logical directions.
6. Diagnostic connection state, MTU/payload handling, sequence logging and reconnect behavior.
7. No cloud/backend.
8. No polished product UI.

Security / pairing

Use platform-supported pairing/encryption where available. Do not invent custom crypto. Surface whether the active link is bonded/encrypted if the platform APIs expose that state.

Background behavior

Implement only what can be justified by current platform APIs. Record foreground/background limitations instead of fabricating parity.

Required validation

Run every software-side validation available in the workspace:

• iOS project generation/build check if the environment supports it
• Wear project Gradle build/test/lint as available
• unit tests for envelope encode/decode + sequence handling
• diff/hygiene check

Do NOT claim the physical transport passes without owner hardware evidence.

Deliverables

Commit implementation to feature/gb-m0-feasibility with an explicit commit message.
Update/create:

• source projects
• diagnostic README/runbook
• exact install/run steps for iPhone + Watch6
• expected evidence format for hardware test
• docs/CURRENT_WORK.md

Final task status must be one of:

• PARTIAL — HARDWARE_REQUIRED if software scaffold/build is ready but physical test has not happened
• BLOCKED with exact blocker
• FAILED with evidence

Never mark R1 PASS from emulator/mock traffic alone.

Final report

Return:

• status
• branch + HEAD
• files changed
• architecture chosen
• exact build/test results
• known limitations
• owner hardware steps
• next recommended PM action

Do not start R2/R3/R4 implementation in this session. Do not merge PR #3.

[PNHD]

CLAUDE REVIEW HANDOFF — RUN ONLY AFTER CODEX COMMIT EXISTS

Session: separate review-only session
Agent: Claude Desktop
Model: Sonnet High
Extended Thinking: ON
Dynamic Workflows: OFF
Sub-agents / agent teams / nested agents: OFF
Role: independent technical reviewer
Workspace: PNHD/GalaxyBridge
Write authority: READ ONLY
Merge authority: NONE

Preconditions

Do not begin substantive review until Codex has committed the R1 implementation on feature/gb-m0-feasibility and reported its HEAD SHA.

Review task

Audit the Codex R1 implementation against:

• issue GB-M0-R1 — iPhone ↔ Watch6 BLE Transport PoC #1 acceptance criteria
• agent_tasks/GB-M0-R1.md
• docs/DECISIONS.md
• docs/M0_TEST_PLAN.md

Verify from the actual diff/source, not from Codex's summary.

Check especially:

1. BLE central/peripheral role correctness on iOS and Wear OS.
2. Whether the chosen Wear OS APIs actually support the implemented server/client behavior.
3. service/characteristic UUID lifecycle and discovery semantics.
4. notification/write-with-response/write-without-response handling.
5. packet framing, sequence, replay/duplicate handling, MTU fragmentation assumptions.
6. reconnect/session reset behavior.
7. background execution claims vs what code actually does.
8. privacy/security: pairing/encryption assumptions, logging, identifiers.
9. build configuration, entitlements/permissions/manifests.
10. whether any test or code path falsely implies hardware PASS.

Evidence

Report exact:

• reviewed branch and HEAD SHA
• files inspected
• commands/tests you actually ran, if any
• findings with severity: BLOCKER, HIGH, MEDIUM, LOW
• exact file/line references
• whether owner physical test packet is sufficient

Final verdict must be exactly one of:

• ACCEPT_SOFTWARE_GATE — HARDWARE_REQUIRED
• CHANGES_REQUIRED
• BLOCKED

Do not edit files. Do not merge. Do not start R2/R3/R4.

[PNHD]

CURRENT EXECUTION SETTINGS — supersedes prior model line

Use the current Codex model picker shown in the owner UI:

• Model: GPT-5.6 Sol
• Effort: High for the first R1 implementation pass
• Speed: Standard
• Sub-agents / parallel agents: OFF

Rationale: GPT-5.6 Sol is the current flagship GPT-5.6 coding model. Keep the first pass at High because R1 is already tightly scoped; escalate a repair/review pass to Sol + Max only if the first implementation exposes hard BLE architecture/platform issues.

The earlier 5.6 Sol line was correct in family but the later PM message recommending GPT-5.3-Codex/xhigh was stale and should be ignored.

PM HANDOFF — REQUIRED

At the end of the task, provide a complete report for the ChatGPT PM containing:

• TASK ID
• FINAL STATUS (PASS, PARTIAL — HARDWARE_REQUIRED, CHANGES_REQUIRED, BLOCKED, or FAILED as applicable)
• REPOSITORY
• BRANCH
• BASELINE HEAD
• FINAL HEAD SHA
• COMMITS CREATED (SHA + message)
• FILES CHANGED (exact paths)
• WHAT WAS IMPLEMENTED
• ARCHITECTURE / TECHNICAL DECISIONS
• COMMANDS ACTUALLY RUN
• BUILD RESULTS
• TEST RESULTS
• UNVALIDATED ITEMS
• HARDWARE EVIDENCE (what was and was not physically tested)
• KNOWN ISSUES / RISKS
• REVIEWER ATTENTION (exact files/functions)
• OWNER ACTION REQUIRED
• NEXT RECOMMENDED TASK (recommend only; do not start it)
• final git status / whether worktree is clean

After producing the PM handoff: STOP. Do not merge, change milestone, or start the next task.

[PNHD]

Agent-skill research is staged separately in draft PR #6 (chore/agent-skills-m0 → feature/gb-m0-feasibility).

Do NOT merge/cherry-pick PR #6 into the active R1 workspace while the current Codex writer session is running. The skill PR will be rebased/reviewed after R1 handoff to preserve the one-writer rule.

The staged contract adds repo-scoped Codex/Claude skills plus the canonical PM handoff template; it does not change R1 application code.

[PNHD]

REUSE AUDIT ADDENDUM — iOS BLE TESTING

For the next R1 implementation/review pass, keep native CoreBluetooth as the production iOS transport. Do not introduce RxSwift/legacy BLE wrappers solely to reduce callback boilerplate.

Approved optional test dependency:

• nordicsemi/IOS-CoreBluetooth-Mock
• version: 1.0.6
• license: BSD-3-Clause
• use only where it materially improves unit/simulator tests of discovery, characteristic IO, fragmentation, disconnect/reconnect and diagnostic UI states.

Mock/simulator results must never promote R1 through the physical hardware gate.

Android/Wear production BLE and iOS production BLE intentionally do not need to use the same library: Wear can use Nordic Android BLE 2.11.0 for its deterministic operation queue while iOS stays on native CoreBluetooth behind a narrow GalaxyBridge adapter.