Summary

• Problem: iMessage DM outbound rows can be misclassified as self-chat when imsg reports the peer handle in both sender and chat_identifier.
• Why it matters: that sends normal DMs down the self-chat echo path, so a cache miss can leak the bot's own outbound reply back into inbound handling and start a loop.
• What changed: the iMessage monitor now uses destination_caller_id when available to distinguish true self-chat from DM false positives, and it preserves that metadata in the parsed payload.
• What did NOT change (scope boundary): no echo-cache redesign, no config changes, and no behavior change for older payloads that do not include destination_caller_id.

Change Type (select all)

• Bug fix
• Feature
• Refactor required for the fix
• Docs
• Security hardening
• Chore/infra

Scope (select all touched areas)

• Gateway / orchestration
• Skills / tool execution
• Auth / tokens
• Memory / storage
• Integrations
• API / contracts
• UI / DX
• CI/CD / infra

Linked Issue/PR

• Closes iMessage plugin: self-echo infinite loop in DM chats (isSelfChat false positive) #61543
• Related iMessage echo loop: DM incorrectly classified as self-chat, bypassing echo filter #59845
• This PR fixes a bug or regression

Root Cause (if applicable)

• Root cause: resolveIMessageInboundDecision() treated sender === chat_identifier as sufficient proof of self-chat, but current imsg payloads can use the peer handle for both fields in a normal DM outbound row.
• Missing detection / guardrail: the monitor ignored destination_caller_id, which is the field that differentiates a real self-chat from this DM false positive.
• Contributing context (if known): the earlier self-chat fixes addressed reflected duplicate rows, but they still depended on the older heuristic for identifying self-chat in the first place.

Regression Test Plan (if applicable)

• Coverage level that should have caught this:
  • Unit test
  • Seam / integration test
  • End-to-end test
  • Existing coverage already sufficient
• Target test or file: extensions/imessage/src/monitor/self-chat-dedupe.test.ts, extensions/imessage/src/monitor.gating.test.ts
• Scenario the test should lock in: is_from_me=true DM rows with sender===chat_identifier but a different destination_caller_id must still take the normal from me drop path, while real self-chat still dispatches.
• Why this is the smallest reliable guardrail: the bug lives entirely inside inbound payload parsing and the inbound-decision branch, so focused tests on those two seams cover the real regression surface.
• Existing test that already covers this (if any): the existing self-chat tests covered the legacy heuristic and true self-chat behavior, but not the DM false-positive shape.
• If no new test is added, why not: N/A

User-visible / Behavior Changes

• OpenClaw no longer treats normal iMessage DM outbound rows as self-chat when imsg provides a different destination_caller_id.
• That keeps those rows on the normal from me drop path and avoids self-echo loops caused by cache misses.

Diagram (if applicable)

Before:
[normal DM outbound row]
  -> sender == chat_identifier
  -> misclassified as self-chat
  -> echo cache miss
  -> dispatched as inbound

After:
[normal DM outbound row]
  -> destination_caller_id != sender
  -> classified as normal DM
  -> dropped as from me

Security Impact (required)

• New permissions/capabilities? No
• Secrets/tokens handling changed? No
• New/changed network calls? No
• Command/tool execution surface changed? No
• Data access scope changed? No
• If any Yes, explain risk + mitigation:

Repro + Verification

Environment

• OS: macOS 15.3 / Darwin 25.3.0
• Runtime/container: Node.js via repo scripts
• Model/provider: N/A
• Integration/channel: iMessage via imsg
• Relevant config (redacted): standard iMessage channel config; no config changes required

Steps

1. Receive a normal iMessage DM where imsg reports an outbound row with the peer handle in both sender and chat_identifier.
2. Let OpenClaw process the is_from_me=true row.
3. Observe whether it takes the self-chat path or the normal from me drop path.

Expected

• The row is recognized as a normal DM outbound event and dropped as from me.
• Real self-chat rows still use the self-chat path.

Actual

• Before this change, the row could be misclassified as self-chat and leak through on an echo-cache miss.
• After this change, the same shape drops as from me when destination_caller_id points at a different local handle.

Evidence

Attach at least one:

• Failing test/log before + passing after
• Trace/log snippets
• Screenshot/recording
• Perf numbers (if relevant)

Human Verification (required)

• Verified scenarios: ran a direct node --import tsx repro against resolveIMessageInboundDecision() for the reported DM false-positive payload shape and for the true self-chat shape; the DM case now drops as from me and the true self-chat case still dispatches.
• Edge cases checked: parser now preserves destination_caller_id; current fallback behavior remains unchanged when that field is absent.
• What you did not verify: the heavier Vitest lanes (pnpm test, pnpm test:extension imessage, and a focused Vitest invocation) stalled after Vitest startup in this environment, so I did not get a completed test summary from those commands.

Review Conversations

• I replied to or resolved every bot review conversation I addressed in this PR.
• I left unresolved only the conversations that still need reviewer or maintainer judgment.

Compatibility / Migration

• Backward compatible? Yes
• Config/env changes? No
• Migration needed? No
• If yes, exact upgrade steps:

Risks and Mitigations

• Risk: older imsg payloads that omit destination_caller_id still rely on the legacy heuristic.

  • Mitigation: the new logic is additive and only tightens classification when the stronger signal is present, so older payload behavior is preserved rather than reinterpreted.

• Risk: this assumes destination_caller_id is the local sending identity for the normal DM false-positive shape.

  • Mitigation: the new regression test locks in the exact reported payload shape, and the change is scoped to is_from_me self-chat classification only.

AI Disclosure

This PR was prepared with AI assistance and reviewed before submission.

Made with Cursor