#11166: fix(whatsapp): detect LID @mentions in self-chat mode

Fixes #8487 #### Summary WhatsApp group @mentions using LID format (e.g. `181891881787642@lid`) are silently ignored when `isSelfChatMode` is true (i.e. the bot's own number is in `allowFrom`). The `else if (hasMentions && isSelfChat)` branch in `isBotMentionedFromTargets()` was completely empty, causing all JID mentions to fall through to text-based matching — which fails because the message body contains LID strings, not phone numbers. lobster-biscuit #### Repro Steps 1. Configure WhatsApp with `allowFrom` containing the bot's own phone number (triggers `isSelfChatMode`) 2. Have another group member @mention the bot 3. WhatsApp sends LID-format JID in `mentionedJids` (e.g. `181891881787642@lid`) 4. `normalizedMentions` correctly resolves the LID to the bot's E164 5. **Bug**: Empty `else if` branch discards this information; `wasMentioned` is `false` #### Root Cause The `else if (hasMentions && isSelfChat)` branch at `src/web/auto-reply/mentions.ts:59-61` was intentionally empty to suppress WhatsApp's auto-included owner JID mentions. However, this also suppressed legitimate @mentions from external group members — including LID-format mentions that `normalizedMentions` correctly resolves to E164. #### Fix Use `msg.senderE164` to distinguish between the owner's own messages (where WhatsApp auto-includes their JID) and external users' messages (where the @mention is intentional): - **Sender IS owner** → suppress JID mentions (preserves original intent) - **Sender is NOT owner** → check `normalizedMentions` against `selfE164` (fixes the bug) #### Behavior Changes - External users @mentioning the bot in a group now correctly triggers the agent session when `isSelfChatMode` is true - Owner's own messages with auto-included JID mentions remain suppressed - Text-based mention patterns (e.g. typing "openclaw") continue to work in all cases #### Codebase and GitHub Search - Searched `isBotMentionedFromTargets`, `isSelfChatMode`, `wasMentioned LID`, `mentionedJids` — no competing PRs - Reviewed PR #5607 (closed, different feature: reply-based implicit mentions) - Verified `normalizeE164` import was already available in `mentions.ts` #### Tests - [x] 3 new unit tests in `mentions.test.ts` (LID from external sender, JID from owner, LID of non-bot user) - [x] Updated integration test to verify owner's messages are still suppressed while external mentions now trigger - [x] All 3 new tests fail before fix, pass after (TDD) - [x] Full suite: 848 test files, 5386 tests pass, 0 failures - [x] `pnpm build` passes - [x] `pnpm check` passes (pre-existing lint error in `memory-lancedb` unrelated to this change) **Sign-Off** - Submitter effort: autonomous - Agent notes: The fix uses `msg.senderE164` (already available on `WebInboundMessage`) to disambiguate the owner's auto-included JID from intentional external mentions. Minimal, targeted fix — no new dependencies, no architectural changes. <!-- greptile_comment --> <h2>Greptile Overview</h2> <h3>Greptile Summary</h3> This change fixes WhatsApp self-chat mode mention detection when group @mentions arrive as LID-format JIDs (e.g. `...@lid`). In `src/web/auto-reply/mentions.ts`, the previously-empty `hasMentions && isSelfChat` branch now distinguishes between messages sent by the owner (where WhatsApp auto-includes the owner’s JID in `mentionedJids` and should remain suppressed) and messages sent by other users (where `normalizedMentions` should be honored to detect bot mentions). Tests were updated accordingly: - `src/web/auto-reply/mentions.test.ts` adds unit coverage for LID mentions from external senders, suppression for owner-sent messages, and non-bot LID mentions. - `src/web/auto-reply.web-auto-reply.supports-always-group-activation-silent-token-preserves.test.ts` extends the existing integration test to ensure owner messages remain suppressed while external JID mentions now trigger in self-chat mode. <h3>Confidence Score: 5/5</h3> - This PR is safe to merge with minimal risk. - The change is narrowly scoped to self-chat mention handling and is covered by new unit and integration tests that exercise the previously-broken scenario (external LID/JID mentions) and the intended suppression behavior for owner-sent messages. The logic relies on existing fields (`senderE164`, `selfE164`) and existing normalization utilities, and no other code paths are affected. - No files require special attention <!-- greptile_other_comments_section --> **Context used:** - Context from `dashboard` - CLAUDE.md ([source](https://app.greptile.com/review/custom-context?memory=fd949e91-5c3a-4ab5-90a1-cbe184fd6ce8)) - Context from `dashboard` - AGENTS.md ([source](https://app.greptile.com/review/custom-context?memory=0d0c8278-ef8e-4d6c-ab21-f5527e322f13)) <!-- /greptile_comment -->