Map agent dependencies and arrange in topological order…

When This Fires

• When designing multi-step cognitive workflows (morning routines, project kickoffs, decision processes)
• When you frequently start tasks only to realize a prerequisite wasn't complete
• When workflow execution under stress produces worse results than under calm — sequencing may be degrading
• Complements Decouple independent sub-behaviors into separate agents… (decouple independent agents) with the positive case: how to properly sequence dependent agents

Common Failure Mode

Sequencing by habit or convenience rather than dependency: "I always check email first because it's easy." But if your priority review doesn't depend on email (it depends on calendar), doing email first delays the dependency chain without contributing to it. Worse, email often derails priority review by introducing reactive tasks. Dependency-based sequencing puts email where it belongs in the chain — after priorities are set, not before.

The Protocol

(1) List all agents/steps in a multi-step workflow. (2) For each, identify its inputs: what information or outputs from other agents does it need to execute properly? (3) Draw the dependency graph: arrows from each agent to the agents that depend on its output. (4) Topologically sort: arrange so each agent appears after all agents it depends on. This is your execution sequence. (5) Where convergence occurs (one agent needs outputs from multiple upstream agents), ensure all upstream agents complete before the downstream agent begins (Map agent dependencies and arrange in topological order… variant on synchronization). (6) Document the sequence explicitly (Under stress, follow the written sequence — intuitive…) — don't rely on intuitive recall of the dependency order.