Recovery Loop 3-Strike

Bound retries so failure handling becomes diagnostic instead of repetitive.

Use When

• a verification step fails and the cause is not yet clear;
• the task is long enough that repeated blind retries are likely;
• you need an explicit escalation path for autonomous or semi-autonomous work.

Core Rules

• Failure one is for learning, not denial.
• Each retry must add new evidence or a narrower hypothesis.
• After two failed passes, stop improvising and write a focused fix plan.
• After the third failure, hand off honestly instead of pretending the loop is still productive.

Workflow