Project Feasibility Reviewer

Problem

Feasibility reviews often fail because industry standards differ, issues lack citations, and cross-project comparisons are hard to keep consistent. This skill turns review rules into a structured workflow: after the user uploads a feasibility report, an estimate sheet or indicator table, and reference files such as regulations, internal checklists, or prior review reports, it extracts review dimensions, weights, checkpoints, and red-line items automatically.

How It Works

The pipeline starts with material parsing: PDF reports use pdf, Word uses docx, Excel uses xlsx, and drawings or scanned pages use OCR/vision analysis to extract tower numbers, equipment models, demolition scope, and other key text. It then reviews each dimension against the extracted rules. Every issue must record the description, cited file and clause, location in the report/drawing/sheet, severity, and category; red-line items trigger a reject decision. compute_score.py handles deterministic scoring, rankings, and red/black lists, and the final deliverable is a .docx review opinion rather than a Markdown file.

Boundaries

It works for single or multi-project reviews in power, municipal, transport, IT, infrastructure, and other industries, provided the relevant review basis is supplied. Review cannot start without rules. Missing reports, estimate sheets, or key annexes are flagged as incomplete materials. For complex spreadsheets, confirm column structure before extraction; the power-distribution attachment module is optional and sector-specific.