bid-review-lite（BID 审查 lite）

Bid Document Review (Lite)

Review bid/tender documents for errors, contradictions, compliance issues, and fraud indicators using AI analysis.

What it does:

• Extracts text + tables from.docx and.doc bid documents
• Cross-references procurement requirements vs bid responses
• Checks pricing consistency, qualification claims, technical parameters
• Identifies contradictions, missing information, expired certificates
• Generates a structured audit report with 🔴/🟡/🟢 risk ratings

Ideal for:

• Bid managers reviewing submissions before deadline
• Procurement officers auditing received bids
• Companies reviewing competitor bids (post-award disclosure)
• Quality assurance on your own tender responses

Quick Start

1. Place your bid documents in a working directory:

- Procurement/tender document (the requirements) - Bid/response document (what was submitted)

1. Tell your agent: Review the bid documents in [directory]. The procurement document is [file1] and the bid response is [file2].
2. The agent will:

- Extract all text and tables - Analyze against the checklist below - Generate a structured report

Text Extraction

For.docx files:

python3 {baseDir}/scripts/extract_text.py input.docx output.txt

For.doc (legacy) files:

python3 {baseDir}/scripts/extract_doc_text.py input.doc output.txt

Review Checklist

1. Pricing & Commercial

• Total price within maximum limit
• Unit prices within per-item limits
• Tax rate calculations correct
• Amount in words matches figures
• No abnormally low pricing (dumping risk)
• Payment terms match requirements

2. Mandatory Requirements (★ items)

• ALL mandatory/starred parameters responded to
• Responses meet or exceed minimums
• Supporting evidence provided for each claim
• No contradictions between different sections

3. Qualification & Eligibility

• Business license valid and matching
• Required certifications in-date
• Performance track record meets minimum
• Credit/reputation checks provided
• Authorization letters (if agent/distributor)

4. Technical Response

• All technical parameters addressed
• Claims supported by test reports/certificates
• Product model matches throughout document
• Standards referenced are current (not withdrawn)
• Delivery timeline realistic vs. claimed

5. Document Integrity

• Bidder name consistent throughout
• Signatures and seals present where required
• Dates logical (no future dates, no pre-bid dates)
• Page numbering sequential
• No template placeholders left unfilled (e.g. [X], [TBD])

6. Common Red Flags

• Identical test results matching nominal values exactly (fabrication indicator)
• Contracts with missing signatures or blank fields
• Expired certificates/qualifications submitted as valid
• Third-party materials without clear authorization
• Inconsistent company names across documents

Report Format

Generate the report in Markdown with this structure:

# Bid Review Report

## Project Info
- Procurement: [name]
- Bidder: [name]
- Date: [date]

## Summary
| Category | Status | Notes |
|----------|--------|-------|
| Pricing  | ✅/⚠️/❌ | ... |
| Mandatory params | ✅/⚠️/❌ | ... |
| ...      | ...    | ...   |

## 🔴 Critical Issues (may cause disqualification)
### 1. [Issue title]
- Location: [where in document]
- Detail: [what's wrong]
- Impact: [consequence]
- Recommendation: [fix]

## 🟡 Medium Issues (affects scoring)
...

## ✅ Positive Findings
...

## Checklist Summary
[Completed checklist with pass/fail for each item]

Dependencies

• Python 3.6+
• python-docx (pip3 install python-docx)
• olefile (pip3 install olefile) — for.doc files only

Limitations (Lite Version)

This lite version covers text-based review only. It does not include:

• Image extraction and visual analysis (certificates, contracts, photos)
• Automated image fraud detection (watermarks, stock photos, expired seals)
• PDF report generation
• Image compression/optimization for API cost savings

For full image review capabilities, see the complete bid-review skill.

Built From Real Experience

This skill was developed from reviewing 3 real bid documents (1,800+ pages, 2,600+ images) in the road maintenance equipment industry. Every checklist item comes from an actual issue found in production reviews.

Issues discovered include: fake test data, stolen web images with visible watermarks, technical parameters contradicting government records by 4 tons, expired certificates submitted as valid, and 30+ pages of copied filler content.