Context Packaging

A handoff is only as good as the context it carries. When HUNTER passes a qualified lead to CLOSER, the context package must contain everything CLOSER needs to start working — without CLOSER having to ask HUNTER follow-up questions. Every follow-up question is a round-trip that costs time, tokens, and momentum.

The context package has four layers. Layer 1 — Summary: a 2-3 sentence overview of what was done and why the handoff is happening. Layer 2 — Structured Data: the payload in the agreed schema (lead details, qualification score, interaction history). Layer 3 — Decision Log: the key decisions made during the upstream agent's work (why this lead was prioritized, what signals triggered qualification, what approaches were considered). Layer 4 — Open Questions: anything the upstream agent could not resolve that the downstream agent should be aware of. Omitting Layer 4 is the most common handoff failure — it forces the receiving agent to discover unknowns through trial and error.

1. Layer 1: Summary Two to three sentences. What happened, what was produced, why it is being handed off. The receiving agent reads this first and knows immediately whether this is relevant to their current priorities.
2. Layer 2: Structured Data The payload — validated against the agreed schema. No free-text substitutes. If the schema requires an ICP score, it gets an ICP score, not "seems like a good fit."
3. Layer 3: Decision Log What was decided, what alternatives were considered, what trade-offs were made. The receiving agent inherits the decision context — they do not need to re-derive the reasoning.
4. Layer 4: Open Questions What the sending agent could not resolve. Explicitly stated, not implied. "Budget not confirmed — prospect mentioned Q2 allocation cycle" is infinitely more useful than silence on the topic.