What problem does it solve? Finished valuations often contain hidden defects: double-counted risks, verdict-first bias, implausible terminal values, and route violations that quietly void the result. This Skill provides a systematic adversarial review that finds those defects and reports them as actionable findings before anyone acts on the number. ## Core Features & Use Cases - Eleven-attack sequence: Runs ordered attacks covering route conformance, bias diagnostics, single-charge-per-risk, growth-reinvestment reconciliation, terminal value plausibility, equity bridge double counts, implied expectations, comparable-set bias, the seven sins of acquisition analysis, narrative traceability, and uncertainty. - Structured findings output: Produces challenge.json with severity-graded findings (id, severity, target_stage, claim, evidence, suggested_fix) plus a prose challenge.md companion, routed to the owning stage for fixes. - Mechanical gate integration: Converts validator output (ERROR/WARN/INFO/SKIP) into findings and invokes sibling toolkits such as dcf-valuation-engine, relative-valuation-toolkit, and cost-of-capital-toolkit for implied-expectation solves and peer statistics. - Use Case: Before an investment committee acts on a DCF, run this review to catch a country risk premium charged alongside an expropriation haircut, quantify the double count per share, and route the fix to the cost-of-capital stage. ## Quick Start Red-team the completed valuation artifacts in this workspace and write challenge.json and challenge.md with severity-graded findings.
