AGENT SKILL

meta-wearables-diagnostics

Diagnose and recover Meta Wearables DAT iOS/Android, native Display, and Display Web App failures, and resolve device and version compatibility — first failing state across registration, permissions, link, session, camera/audio/Display, transport, firmware, and companion, with a redacted diagnostic bundle and a validated compatibility packet. Use when a DAT app cannot reach the glasses, registration or a session fails, an update-required or device-unavailable error appears, firmware or Meta AI versions drift, or someone asks whether a named product such as Gen 2 or Gen 3 supports a capability.

License 许可
MIT
最近更新
2026-11-10
Stars
★ 0
Forks
0

01 前言 Front Matter

name
meta-wearables-diagnostics
description
Diagnose and recover Meta Wearables DAT iOS/Android, native Display, and Display Web App failures, and resolve device and version compatibility — first failing state across registration, permissions, link, session, camera/audio/Display, transport, firmware, and companion, with a redacted diagnostic bundle and a validated compatibility packet. Use when a DAT app cannot reach the glasses, registration or a session fails, an update-required or device-unavailable error appears, firmware or Meta AI versions drift, or someone asks whether a named product such as Gen 2 or Gen 3 supports a capability.
allowed-tools
Read, Grep, Glob, Bash(uv run *)
last_verified
1787356800000

02 资源文件 5

03 SKILL.md 正文

Meta Wearables diagnostics

Find the first failing state, apply the least invasive documented recovery, and decide compatibility from an observed version tuple, never from a product label. Diagnosis is observation: it is not permission to change the Meta AI app, the glasses, the account, registration, or a release channel.

Apply the shared rules — evidence ladder, device labels, secrets, claims.

Read before acting

Workflow

  1. Freeze the tuple. Platform, app build, SDK package or artifact and tag, phone OS, Meta AI version, glasses firmware, on-glasses DAT app version, product label, runtime device type, account mode (Developer Mode or release channel), and the one operation being attempted. Mark gaps to-verify.
  2. Collect the baseline. If a DAT live-debugging MCP server is connected, discover its tools and use them read-only. If none is connected, fall back to platform logs (the Xcode console or Console.app filtered to the app's os_log subsystem; adb logcat --pid=$(adb shell pidof <application.id>)), the app's own logged DAT registration, device, link, session, and typed-error transitions, a MockDevice reproduction of the same operation, and a hand-built redacted diagnostic bundle. The diagnostic contract lists the steps.
  3. Walk the path in order: initialization and configuration -> callback -> registration -> permission -> device eligibility and compatibility -> link -> session -> camera/audio/Display -> transport -> thermal/power. The first transition that fails owns the diagnosis; record later failures as symptoms. Use the platform's own typed errors; never translate iOS and Android symbols.
  4. Reproduce once, narrowly. One operation, one run ID; stop at the earliest actionable boundary.
  5. Recover with the least invasive documented step, then re-run the same operation under a new run ID. Never auto-retry through thermal emergency, critical battery, power shutdown, doff/hinge closure, or an unknown protocol error. A firmware update, DAT-app update, or re-pair counts as successful only when the post-action state is reported and the original operation completes.
  6. Resolve compatibility per capability. Compare each tuple member with the official version-dependency surface and label it source, observed, access-gated, conflict, or unknown. Exercise each requested capability on its own, including a negative path, and give a fallback for every unresolved one.
  7. Validate the packet. Fill the compatibility packet template and run the validator below before handing off.

Running the validator

$SKILL below is this skill's directory — the folder holding this SKILL.md. Find it by name in your tool's skills folder (some tools prefix it, e.g. meta-wearables-ops__meta-wearables-diagnostics).

uv run $SKILL/scripts/validate_compatibility_packet.py <packet.yaml>

A completed packet needs connected-tuple, physical-capability, and release evidence rows. A Gen 3 packet cannot close without COMP-GEN3-01 physical and release evidence.

Hard boundaries

  • Do not register, unregister, change permissions, re-pair, update firmware, or switch a channel during diagnosis unless the user authorizes that operation.
  • Do not diagnose from a generic "not connected" message when the first failing state can be observed.
  • A debug-server or log result is app-visible evidence through the DAT boundary, not introspection of the Meta AI companion.
  • A GitHub issue or forum report is a troubleshooting signal, never a compatibility guarantee. Label it separately.
  • Developer Mode success is not release-channel evidence. Re-test with the signed build, tester account, and exact firmware before calling a route ready.

Required output

  1. Frozen tuple with each field labeled source, observed, access-gated, or to-verify, plus the evidence level reached.
  2. Baseline source used (MCP server, or the named fallback steps) and the ordered event trace with the exact first failing state.
  3. Likely owner: app, SDK/config, Meta AI boundary, device/firmware, transport, audio, thermal/power, or release channel.
  4. Recovery attempted, post-action state, and whether the original operation then completed.
  5. Capability-by-capability status (current, to-verify, device-gated, source-conflict, unsupported) with a fallback for each gap, and explicit Gen 2 / Gen 3 treatment.
  6. Evidence rows (DBG-*, OPS-*, COMP-*) with redaction status and the validator receipt.
  7. Verdict — ready, ready-for-device, blocked, source-conflict, or to-verify — with the proof still missing and the next refresh trigger.

Related skills

04 同仓库更多 Skills