Construction Cost Audit AI Assistant

Problem it addresses

Construction cost audits often stall when settlement data, quantity bills, claim evidence, rate tables, and contract clauses are scattered across files and reviewers use different assumptions. This skill structures that work as a multi-role audit workflow, assigning settlement, bill-of-quantities, claim, and metrics tasks to specialized agents before producing a traceable summary.

How it works

• Context first: the lead agent extracts project stage, deliverable expectations, and constraints, then asks structured questions when contracts, drawings, or settlement sheets are missing.
• Role split: settlement, bill-of-quantities, claim, metrics, and cost-audit roles each handle their scope, reducing cross-talk and duplicated assumptions.
• Evidence-backed calculation: numeric work uses calc_engine.py and checks clause references in references/standards/规范数据库.json; local price data must be supplied or imported into local_rates.json.
• Output gate: final answers pass through gate.py, which blocks vague wording such as “probably” or “maybe” and specific numbers lacking clause support, and can generate an xlsx comparison sheet and a Markdown audit report.

Boundaries

It is an audit-assistance tool, not an official pricing authority. Its numeric outputs are AI reference values and must be verified against local published price information, current quota standards, fee rules, and real project documents before being used for settlement, bidding control prices, or legal conclusions.