fix(agents): recursively decompose proof obligations

[FluffyAIcode]

Summary

• replace general claim auditing with recursive_proof_decomposition_v2
• focus exclusively on the central mathematical claim and proof chain
• attack unknown, unsolved, or impossible as stopping claims rather than answers
• recursively decompose broad unresolved nodes into smaller proof obligations
• require argument, strongest counterargument, dependencies, and status per node
• terminate decomposition only at explicitly derived leaves or precisely stated open lemmas
• report the smallest unresolved frontier and next adversarial step
• explicitly ignore prizes, money, prestige, style, and other proof-irrelevant noise

Runtime reset

All old REPL processes, active Prefill jobs, Primary hot KV, and allens cache state were explicitly cleared before validating this protocol.

Tests

• deterministic recursive protocol and full-context invariants — 17 targeted tests pass
• full platform-neutral suite — 867 passed, 1 skipped
• Python compile and diff checks pass

Made with Cursor

[cursor]

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.