Agent priority orderings must be context-specific — which agent wins depends on the time, capacity state, and situation
Build priority orderings that are context-specific rather than global, defining which agent takes precedence during specific time blocks, capacity states, or situational contexts rather than attempting universal rankings.
Why This Is a Rule
Global priority orderings ("health always beats productivity which always beats social") fail because the right priority genuinely varies by context. During a work deadline, productivity should outrank social. During a friend's crisis, social should outrank productivity. During illness, health should outrank everything. A global ranking that says "health > productivity > social" always produces wrong answers in contexts where the ranking should be different.
Context-specific orderings define which agent takes precedence within specific conditions: Time-based: "9 AM-12 PM: deep work agent leads. 12-1 PM: health agent leads (lunch/movement). 1-5 PM: productivity agent leads." Capacity-based: "When energy is high: productivity agent leads. When energy is depleted: recovery agent leads." Situational: "During meetings: social/communication agent leads. During solo work: deep focus agent leads. During crisis: the relevant domain agent leads regardless of time."
This creates a routing table, not a universal hierarchy. The routing table determines which agent is accountable (Assign exactly one agent as accountable for each decision — consulted agents advise, but only one has final authority) based on current context, producing different but pre-determined orderings for different situations.
When This Fires
- When designing the priority hierarchy (Resolve inter-agent conflicts with documented priority hierarchies — case-by-case deliberation defeats the purpose of automation) and finding that a single ranking doesn't cover all situations
- When the "right" priority genuinely varies by time of day, energy level, or situation
- When a global priority ranking keeps producing wrong answers in specific contexts
- When building a complete multi-agent coordination architecture
Common Failure Mode
Attempting universal rankings: "My health is my #1 priority." Always? During a job interview? During a once-in-a-career opportunity? During a team emergency? Universal rankings sound principled but produce rigid behavior that's wrong in predictable contexts. Context-specific rankings are more complex to design but produce correct behavior across varied situations.
The Protocol
(1) Identify the contexts in which your agents' priorities genuinely vary. Common context dimensions: time of day, energy/capacity state, social situation, deadline proximity, and life phase. (2) For each context, define the priority ordering: which agent takes precedence? (3) Format as a routing table: "During [context] → [agent] leads, [other agents] are consulted." (4) Define context transition triggers: how do you know which context you're in? Use observable criteria (Apply the camera test to triggers — if a camera can't detect the exact firing moment, the trigger is too vague), not subjective assessment. "After 12 PM" not "when I feel like taking a break." (5) When agents conflict, consult the routing table: what context am I in right now? The table resolves the conflict without real-time deliberation.