Problem

When an agent is @-mentioned in a Slack thread whose parent is a third-party bot (Intercom escalations, GitHub/Sentry alerts, webhook posts), the parent message — often the entire reason for the ping — never reaches the model. The agent then has to ask the user for clarification or manually fetch the thread via the Slack API.

Root cause

_fetch_thread_context() in gateway/platforms/slack.py filtered every message carrying a bot_id or subtype: bot_message:

# Exclude our own bot messages to avoid circular context.
if msg.get("bot_id") or msg.get("subtype") == "bot_message":
    continue

The comment says "our own bot messages" but the predicate matches all bots.

Fix

Match the filter against the resolved bot_uid (already computed a few lines above). Our own replies stay excluded (no circular context); every other bot's messages flow through as normal thread context.

If bot_uid is unresolved, fall back to the old behavior (drop all bot messages) so we can't accidentally feed our own output back in.

is_bot_msg = bool(msg.get("bot_id")) or msg.get("subtype") == "bot_message"
if is_bot_msg and (not bot_uid or msg.get("user") == bot_uid):
    continue

Testing

Locally reproduced with an Intercom-style bot-authored thread parent:

• Before: agent saw only the user's [investigate] ping, no parent.
• After: [Thread context —] block includes the bot parent verbatim.