Claude's Free Finance Plugins vs 5 Planted Errors (Free File)

FREE FILE (the fake books, all 3 prompts, the install commands, the answer key): https://drive.google.com/file/d/1T1kO... Anthropic's free finance and data plugins for Claude (github.com/anthropics/knowledge-work-plugins), tested with one prompt each on a fictional company's September books. I planted 5 errors in a bank reconciliation plus one trap, and scored every answer against my answer key. Results: the bank reconciliation found all 5 planted items, both adjusted balances came to 265,741.86 with a zero difference, three correct draft journal entries, and it left the 4 identical freight payments alone. It also caught a real issue I didn't plant. The variance analysis got every price and volume number right, but gave reasons that are not in the data. The data plugin found the same errors, and then wrote 265,471.86 instead of 265,741.86 in its summary. Chapters 0:00 Claude's free finance plugins vs 5 planted errors 0:58 What the plugins are + install in two commands 1:57 Test 1: bank reconciliation (5 planted errors) 3:56 Test 2: variance analysis (price vs volume) 5:03 Test 3: data plugin profiles the ledger 5:54 What it got wrong (and how to check) 7:20 Scorecard, your own books + quiz Install (Claude Code, in the folder with your files; --scope local = only that folder) claude plugin marketplace add anthropics/knowledge-work-plugins --scope local claude plugin install finance@knowledge-work-plugins --scope local claude plugin install data@knowledge-work-plugins --scope local claude Prompts used (each run once on Oct 21, 2026, Claude Code, Opus 5.5; finance plugin 1.3.0, data plugin 1.1.0) Test 1: /finance:reconciliation Cash account 1010, September 2026. The general ledger is NW_GL_Cash_1010_Sep2026.csv and the bank statement is NW_Bank_Statement_Sep2026.csv. Find every reconciling item, show the adjusted balances, and save the reconciliation as Bank_Rec_Sep2026.xlsx. Test 2: /finance:variance-analysis Revenue and expenses, September 2026 vs budget. Use NW_PnL_Sep2026.csv and NW_Sales_Drivers_Sep2026.csv. Split the revenue variances into price and volume, list every material variance with its likely cause, and save it as Variance_Sep2026.xlsx. Test 3: /data:explore-data NW_GL_Cash_1010_Sep2026.csv. Profile this ledger export and flag anything that looks wrong. Scorecard (0-2 points each for numbers right, complete, no false flags or made-up reasons, accounting judgement, easy to audit): bank reconciliation 10/10, variance analysis 9/10, data profile 9/10. Honest notes: I planted the errors and knew the answers; on real books you won't. The plugins draft, a qualified person reviews and signs off. Not financial advice. Plugins need a paid Claude plan (Pro, Max, Team or Enterprise). My demo folder had a one-line note telling Claude not to automate Excel on this PC, so the saved workbooks have formulas but no stored values until you open them in Excel. Results vary from run to run. Screens are the real recording; on-screen quotes are Claude's own words from the session. Quiz answers are in the pinned comment. Next: Budget vs Actual in Power BI, built with Claude:    • Budget vs Actual in Power BI: Build It wit...   Northwind Outfitters is a fictional company; all numbers are practice data in US dollars. #Claude #ClaudeCode #Accounting #BankReconciliation #Excel