Boo AI Technical Bid Review System

Problem

Technical bid documents often contain high-risk errors before submission, such as wrong project names, accidental inclusion of entity names, cross-type solution mismatches, or citations to abolished standards. Manual line-by-line review is labor-intensive and can miss contradictions across sections. This skill makes the review process rule-based and traceable, rather than generating bid content.

How It Works

It runs sequentially from Stage 0 through Stage 5: environment check, feature fingerprint, agent scanning, mapping, guidelines, chunk-level mechanical scan, sentence-level semantic reading, dimension scoring, a 10-dimension matrix, three-expert semantic reasoning, narrative synthesis, and adversarial validation. Key constraints include A+/A/B/C risk levels, A+ as a veto, verbatim original-text citations, and chapter-section-paragraph location IDs. The final verdict is based on the number of A+, A, and B findings, producing not recommended, revise before resubmission, or recommended for submission.

Boundaries

It operates as an independent review system and produces its own review basis without relying on external intermediate files. It is best suited for structured technical bid review, risk localization, and submission decisions. If the document language, format, or domain rules exceed its 17 rules and risk model, human judgment is still required.