chat-group-behavior（聊天群行为）

Chat Group Behavior

This skill covers the full lifecycle of group chat participation: joining, behaving well while active, and leaving cleanly.

Prerequisites

channels.<platform>.groupPolicy, groupAllowFrom, and groups.*.requireMention are set so that only the owner can wake you in any group by default. Trigger phrases are defined in agents.list[main].groupChat.mentionPatterns. This applies to WhatsApp, Signal, Telegram, and other supported platforms.

Joining a Group

When the owner @mentions you in a group chat, follow these steps in order.

Step 1 — Capture context

From the inbound message metadata: platform (e.g. whatsapp, signal), group ID (the chat_id / from field), and duration (parse from the owner's message — default 30 minutes if unspecified).

Step 2 — Open the gate

Patch channels.<platform>.groups.<group_id>.requireMention to false in openclaw.json. The gateway file-watches the config and hot-reloads automatically — no restart needed.

Step 2b — Open the sender filter ⚠️ Critical — easy to miss

groupAllowFrom controls not just *who can wake you* in a group, but also *which senders' messages are delivered to you* once active. If it's set to only the owner's number (the default), messages from other group members will be silently dropped even with requireMention: false.

Fix: patch channels.<platform>.groupAllowFrom to ["*"] so all group members' messages are delivered.

This is safe because groupPolicy: "allowlist" still restricts *which groups* you are active in — opening the sender filter only affects groups you've explicitly joined.

Verify with openclaw doctor — config changes must pass schema validation.

Step 3 — Signal before going quiet

Before doing any work (research, lookups, etc.) triggered by a group message, send a brief acknowledgment first: *"On it, give me a moment 👁️"* — then go do the work. Silence looks like a crash.

Step 4 — Introduce yourself

Say hi, introduce yourself briefly if it's a new group. Participate like a human — use judgment about when to contribute. Do not respond to every message.

Step 5 — Set a closing cron job

Create a one-shot cron job (kind: "at") for the parsed duration. When it fires: patch requireMention back to true for this group in openclaw.json, then send a brief goodbye. The gateway hot-reloads and the gate closes — no owner action needed.

Step 6 — Confirm to owner

Tell the owner the window is open and when it closes: _"Joined! I'll be here for 60 minutes — going quiet at 4:19 PM."_

Behavior While Active

• One response per message — don't triple-tap with different reactions
• Signal before working — acknowledge first, deliver second (see Step 3)
• Don't dominate — respond when you add value; stay quiet for casual banter between humans
• Be present, not performative — contribute naturally, like a human participant

Leaving a Group

Timed close (automatic): When the closing cron fires: patch requireMention: true back for this group in openclaw.json, send a brief goodbye to the group. Gateway hot-reloads.

Early close (owner request): Owner says @<agent> that's enough / close → patch requireMention: true immediately, cancel the cron, say goodbye.

Extend (owner request): Owner says @<agent> stay for another X mins → cancel existing cron, set a new one.

Troubleshooting

Messages from group members not reaching me: Check groupAllowFrom — if it's restricted to the owner's number, non-owner messages are silently filtered. Set to ["*"] (see Step 2b).

"thinking/redacted_thinking blocks cannot be modified" errors: This happens when the session has reasoning tokens in its history and a subsequent API call tries to re-process them. Fix: clear the session transcript file (find it via sessions.json in the agent sessions directory). The session will restart clean on the next message.

Notes

• Group ID is in inbound message metadata (chat_id / from field)
• Config changes hot-reload — no gateway restart or Docker restart needed
• When closing, set requireMention: true for the specific group entry (or remove it to fall back to the wildcard default)
• groupAllowFrom: ["*"] combined with groupPolicy: "allowlist" is the correct secure pattern — open sender filter, restricted group allowlist