fix(slack): check Block Kit blocks for @mentions, not just flat text

[AlexFucuson9]

Problem

Mention detection only checked the flat text field for @mentions. Block Kit messages that have the @mention only in blocks (e.g. rich_text sections with user elements) were silently dropped.

# Before — only checks flat text
routing_text = original_text or ""
is_mentioned = bool(
    (bot_uid and f"<@{bot_uid}>" in routing_text)
    or self._slack_message_matches_mention_patterns(routing_text)
)

# After — checks both flat text and blocks text
blocks = event.get("blocks")
blocks_text = _extract_text_from_slack_blocks(blocks) if blocks else ""
mention_text = (original_text or "") + ("\n" + blocks_text if blocks_text else "")
is_mentioned = bool(
    (bot_uid and f"<@{bot_uid}>" in mention_text)
    or self._slack_message_matches_mention_patterns(mention_text)
)

Files changed

File	Change
plugins/platforms/slack/adapter.py	Extract blocks text and include in mention check (+8/-3 lines)

Fixes #52387

[tonydwb]

Code Review Summary

Verdict: Approved

Small, focused fix for Slack Block Kit mention detection. The change correctly extracts text from blocks and includes it in the mention check. Note: PR #52390 addresses the same issue (#52387) with a more comprehensive approach including helper functions and additional test coverage — the author/reviewer should reconcile these.

Looks Good

• Minimal diff (8 additions, 3 deletions)
• Correct approach: merge blocks text into mention detection
• References the existing _extract_text_from_slack_blocks helper

---

Reviewed by Hermes Agent

[teknium1]

Thank you for the focused Slack investigation. The current main code still has the reported Block Kit mention gap, but this patch needs rework before it can safely address it.

Problems

• The patch changes only channel routing. The earlier allow_bots: mentions filter still reads flat event["text"] and returns at plugins/platforms/slack/adapter.py:2607-2610, so the issue's bot-message reproduction remains dropped.
• _extract_text_from_slack_blocks() recursively includes rich_text_quote content (plugins/platforms/slack/adapter.py:173-174) and renders user elements as mention tokens (:145-146). Feeding all extracted text into routing would violate the existing quoted-mention safety contract in tests/gateway/test_slack.py:1687-1721.
• The PR adds no regression tests for either mention gate or for preserving quoted-content isolation.

Suggested changes

• Share a detection helper between the two gates, collecting only non-quoted Block Kit user mentions.
• Add integration tests for a direct Block Kit mention, allow_bots: mentions, and quoted-only mention rejection.

Automated hermes-sweeper review.

[teknium1]

_extract_text_from_slack_blocks() includes nested rich_text_quote content, so this makes a quoted/forwarded <@bot> satisfy routing. Current main deliberately rejects that case in tests/gateway/test_slack.py:1687-1721; use a detector that collects only non-quoted user elements instead.

[teknium1]

Closing as superseded by #69316 (merged): counting mentions in ALL block text (incl. quoted/forwarded) risks response hijack; the landed fix (#52390 base) only counts non-quoted mentions.

Thanks for digging into this — the consolidated fix stands on the cluster's collective analysis, and your work is credited in #69316's summary.