human-in-the-loop

1# Human In The Loop
2 
3The primitive that treats humans as **durable network participants** —
4attention surfaces, queues, decision records, routing semantics — not
5as ad-hoc chat receivers.
6 
7**Autonomy is not the absence of humans; it is knowing when human
8judgment is needed and making that handoff crisp.**
9 
10## Use this when
11 
12- A slice closeout needs classifying: auto-continue, human gate, or park
13- A real decision needs to land in front of a human (usage limits, provider auth, roadmap tradeoff, product-intent ambiguity)
14- Designing a human queue/dashboard surface
15- Returning a hot potato to orchestration after human approval
16 
17## Don't use this when
18 
19- The slice closeout is clean and `PROGRESS.md` already names the next safe slice. **Default RSI conveyor continues; do NOT manufacture a human gate.**
20- The escalation is just a status update. Humans are participants for *decisions*, not narration.
21- The next owner is another agent. Use queue-handoff, not human-in-the-loop.
22 
23## The 3-class closeout classification
24 
25In a productized daemon-backed version, closeout classifies the next
26step BEFORE touching the human queue:
27 
28| Class | When | Action |
29|---|---|---|
30| **auto-continue** | Slice closes cleanly, next named slice in workstream plan | Mark closed; create next-owner qitem from plan |
31…