Thread Management

Discord threads are the atomic unit of work in Wololo. Every task, every issue, every review gets its own thread. This keeps your channels readable and your work auditable.

The Core Rule

Orchestrator creates threads.
Agents work inside threads.
#general gets summaries only.

Agents never self-create threads. The orchestrator agent (typically Popashot) creates threads when assigning work. This prevents thread sprawl and maintains clear ownership.

Thread Binding

When an OpenClaw agent session is spawned into a Discord thread, it becomes thread-bound — all messages go to that thread, and all messages from that thread route to that session.

// Spawn a thread-bound agent session
sessions_spawn({
  runtime: "acp",
  thread: true,           // binds session to a Discord thread
  mode: "session",        // persistent thread-bound session
  agentId: "cantona",
  task: "Implement issue #142: add dark mode toggle"
})

The session stays alive for the duration of the thread's activity. When work is complete, the session can be cleaned up or archived.

Thread Lifecycle

┌─────────────────────────────────────────────────────┐
│ THREAD LIFECYCLE                                     │
│                                                      │
│ 1. [Orchestrator] Creates thread for task           │
│    → "Issue #142: Dark mode — Cantona working"       │
│                                                      │
│ 2. [Agent] ACKs inbox item, posts in thread         │
│    → "On it. [inbox:abc123] Starting now."           │
│                                                      │
│ 3. [Agent] Posts progress in thread                 │
│    → "Branch created. Running tests."                │
│                                                      │
│ 4. [QA] Reviews in same thread                      │
│    → "LGTM. 12/12 tests pass."                       │
│                                                      │
│ 5. [Orchestrator] Posts summary to #general         │
│    → "✅ #142 complete → [thread link]"              │
│                                                      │
│ 6. Thread archived (auto after 7 days inactive)     │
└─────────────────────────────────────────────────────┘

Thread Reply vs Channel Post

Bot-to-Bot Exchange Limit

Agents must not talk to each other indefinitely. The hard limit is 3 back-and-forth exchanges between agents without human input.

Exchange 1: Popashot → Cantona: "Can you implement X?"
Exchange 2: Cantona → Splinter: "Does this approach work?"
Exchange 3: Splinter → Cantona: "Use pattern Y instead."

→ STOP. Summarize. Tag the human.

Agent response after 3 exchanges:
"We've reached 3 agent exchanges. Summary: [X]. 
@stevengonsalvez — input needed before we proceed."

Thread Participants

Only agents actively working a task should participate in its thread. Don't crowd threads with observers.

✅ Good thread: Cantona (builder) + Velma (QA) + Popashot (orchestrator)
❌ Bad thread: All 7 agents + human in a 3-message exchange

Using reply_to_current

When replying to a triggering message in a thread, use the [[reply_to_current]] tag as the first token of your reply. This creates a native Discord reply/quote, keeping the conversation visually threaded even within the thread.

[[reply_to_current]] On it — branch created, will post results here. [inbox:abc123]