finance-rumor closed-book skill

To verify listed-company rumors without hallucinated confirmation, use an evidence-tiered assessment that separates rumor claims, verified facts, and market signals into distinct confidence tiers. This enforces closed-book constraints to prevent treating community discussion as proof.

Evidence tiering in finance rumor verification is the process of ranking source reliability, such as separating A1/A2 verified facts from C-only community signals. It ensures that outputs respect source confidence tiers and apply abstain conditions when evidence is weak.

To assess market sensitivity for merger rumor event packs, evaluate the trading impact alongside separated rumor claims and verified facts. Apply market-sensitivity-aware abstention to lower confidence or abstain entirely when trading impact cannot be confirmed.

You should abstain from making risk assessments on company rumors when evidence is weak, source confidence is limited to C-only signals, or trading impact cannot be confirmed. Closed-book execution requires abstaining rather than hallucinating confirmation.

Closed-book execution does not require external corpus, memory, or rubric judge-bundle inputs for rumor verification. It evaluates claims strictly against the provided domain context and source tiers within the event pack itself.