Add verification checkpoints at every coupling point where…

When This Fires

• When mapping multi-step processes to identify error propagation risk
• When an error in the final output is difficult to trace back to its source — the pipeline lacks intermediate verification
• When designing new multi-step workflows or automations
• Complements Place pre-flight checks at the prep/execution boundary… (pre-flight positioning) with the specific criterion for where to place verifications within a pipeline

Common Failure Mode

Trusting upstream processes because they "usually work": "Step 2 always gets correct input from Step 1, so we don't check." This trust is justified until the day Step 1 produces an anomalous output — and then every downstream step amplifies the anomaly because no verification caught it. The coupling point verification is insurance: rarely needed, but when needed, it prevents cascade amplification that's far more expensive than the verification cost.

The Protocol

(1) Map your multi-step process as a pipeline: what's the output of each step, and what consumes that output? (2) Identify coupling points: where does one step's output feed directly into the next step without any verification? These are your cascade risk points. (3) At each coupling point, insert a verification: does the output from the upstream step meet the input requirements of the downstream step? (4) The verification should be independent — not performed by the same process that produced the output (same-process verification has confirmation bias). (5) The verification doesn't need to be exhaustive — a quick sanity check that catches the most common upstream errors is sufficient. The goal is cascade interruption, not perfection at each checkpoint.