Foundry automation benchmark

RPS Interpretation

Interpretive layer for strategy dynamics, regret, equilibrium checks, and exploitability warnings. A Scylla specification is completed into Homer, Athena, Prometheus, audit, and review artifacts; the benchmark checks whether generated structure is executable enough to admit.

Local Sections

input

Scylla spec

  • entitiespresent
  • decisionspresent
  • uncertaintiespresent
construction

Homer program

  • contextsrequired
  • predicatesrequired
  • artifact contractrequired
geometry

Athena plan

  • overlapsrequired
  • restriction mapsrequired
  • decision surfacesrequired
review

Prometheus bundle

  • world modelrequired
  • gluing auditrequired
  • Scylla questionsrequired

Restriction And Gluing

OverlapContextsStatusDecision Surface
spec_programScylla spec / Homer programgluesdomain objects become local contexts
program_planHomer program / Athena planqualifiedoverlap predicates must be explicit
plan_bundleAthena plan / Prometheus bundlereviewartifact contracts must name promotion and obstruction behavior
The benchmark rewards generated foundries that compute local contexts and restriction maps, not just fluent descriptions of a workflow.