Assumptions modeled as provable assignments for an operator:
((assignment "MatchOnEffect"                                 %%Rule Consequent%%
    (?ws ?fluent ?listofpairs ?state ?preceding-state ) %% Plan assignment statement%%
    ("Assume" ?state (?prop ?value)))                         %%Statement of assumption%%
      ⇐ ...                                                                  %%Rule antecedents%%
(Output ?ws ?prop ?value)
(Selectable ?ws ?prop))

Contingencies explict part of the Redux model.
For each appropriate operator, prove that each possible output that is one of the possible-output-values and not the one that was selectable is a contingency.
When such a contingency is asserted, Redux dependencies ensure that this part of the plan becomes invalid and goals need to be re-satisfied by re-planning.
GOOP program goal-regression algorithm with resequencing to get a planner.
Start with a fully instantiated goal Ψ and propagate the values to each (?prop ?value) to create a process instance with no workflow constructs.