Smart Bid Document Review Expert

Problem Addressed

Bid review often involves hundreds of pages of docx/doc/pdf files, where missing formatting, absent signatures, weak scoring responses, missing qualification evidence, negative technical deviations, inconsistent price amounts, or duplicated technical content can lead to rejection or low scores. bid-checker-pro is not a bid-writing tool; it extracts requirements from the tender document and checks them against the actual bid submission, producing a traceable audit report.

How It Works

The skill expects a tender file, a bid file, and a project name, and defaults to a full review. It first loads experience rules and audit configuration, then checks local RapidOCR availability. Document conversion tries MS Word COM, WPS COM, LibreOffice, and Aspose.Words before falling back to lossy extraction, so tables and layout are preserved when possible.

Key checks include:

• Tender requirement extraction: section structure, mandatory items, signatures, scoring tables, qualification requirements, and technical/commercial deviation tables.
• Bid indexing: chapter tree, summaries, and keyword index to reduce large-document context pressure.
• Starred items and hidden rejection clauses: detects ★/⭐/☆, rejection keywords, and adds LLM review to catch clauses without stars.
• Text and amount checks: typos, punctuation, uppercase/lowercase amount consistency, and company/project name consistency.
• Deviation table checks: technical and commercial deviations are checked row by row, including hidden negative deviations.
• Qualification and visual checks: verifies each qualification requirement against the bid text and flags image, layout, and signature risks.
• Report gatekeeping: the final Markdown report must include 11 chapters; missing chapters trigger rework.

Boundaries

It is useful for pre-submission compliance review and scoring-risk triage, but it cannot replace legal, finance, or procurement experts for final decisions. If conversion degrades to text_only, table, pricing, and signature conclusions should be manually reviewed. Price scoring depends on the competitor benchmark price provided by the user, and unavailable sub-capabilities are marked as degraded rather than silently skipped.