transaction-processing
Skill2 starsChanged 7 months agoMITView on GitHub
Processes daily financial transactions including invoices, expenses, receipts, credit card charges, and reimbursements with proper categorization, duplicate detection, anomaly flagging, and general ledger posting preparation. Use when the user mentions processing transactions, categorizing expenses, entering receipts into the books, uploading CSV exports from Stripe or QuickBooks, or asks about transaction reconciliation and bookkeeping automation.
In this repository35 other files it gives its agents.
- banking-relationship-managementSkill
- bank-reconciliationSkill
- cash-forecastingSkill
- cash-monitoringSkill
- working-capital-optimizationSkill
- cfo-orchestratorSkill
- accounts-payable-managementSkill
- accounts-receivable-managementSkill
- ledger-managementSkill
- payroll-processingSkill
- subscription-managementSkill
- budget-creation-managementSkill
- business-case-modelingSkill
- forecast-accuracy-trackingSkill
- revenue-forecastingSkill
- scenario-planningSkill
- unit-economics-analysisSkill
- cap-table-managementSkill
- data-room-managementSkill
- fundraising-financialsSkill
- fundraising-process-managementSkill
- investor-relationsSkill
- term-sheet-analysisSkill
- audit-preparationSkill
- board-reportingSkill
- financial-statement-generationSkill
- internal-controls-designSkill
- monthly-close-processSkill
- tax-compliance-managementSkill
- cost-optimizationSkill
- headcount-and-comp-planningSkill
- pricing-strategy-advisorySkill
- profitability-analysisSkill
- risk-managementSkill
- strategic-initiative-modelingSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 60
Deduplicate: Check for duplicates against the existing transaction log.
- Distinctive: line 99
All transactions in the source data are parsed, deduplicated, and categorized
- Distinctive: line 102
Every transaction has a confidence score (HIGH, MEDIUM, or LOW)
- Distinctive: line 104
No transaction is auto-posted to the ledger without human review when flags exist
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Transaction Processing
Purpose
Every financial transaction that hits the business — invoices received, expense reports, credit card charges, reimbursements — must be categorized, deduplicated, and flagged for anomalies before it reaches the general ledger. This skill ensures every dollar in and out is accounted for, properly categorized, and ready for posting, reducing bookkeeping errors and providing clean financial data for reporting.
When to Use
- "Process these transactions / receipts / invoices"
- "Categorize these expenses"
- "Enter these into the books"
- Upload of receipt images, CSV exports from bank/Stripe/QuickBooks, or raw transaction lists.
- "Reconcile and categorize this month's transactions"
Inputs Required
You must gather or receive:
- Transaction source data — CSV export, bank feed, API dump, or manual list. Must include: date, amount, counterparty, description.
- Chart of accounts — if not already known, load via
ledger-managementskill. - Previous period categorization — optional, for consistency checking.
Quick Reference
| Concept | What It Does | Why It Matters |
| Auto-categorization | Maps transaction descriptions to COA categories using rules | Ensures every transaction has a home without manual effort |
| Deduplication | Scans for duplicate amounts/vendor/date combinations | Prevents double-counting, the most common bookkeeping error |
| Anomaly detection | Flags outlier amounts, missing fields, round numbers | Catches errors and potential fraud before ledger posting |
| Confidence scoring | HIGH / MEDIUM / LOW per categorization | Surfaces items that need human review |
| Multi-currency handling | Records original + USD-equivalent | Required for accurate P&L and cash reporting |
Procedure
- Parse the transaction file: detect format (CSV, JSON, PDF table). Extract: date, amount, counterparty, description, payment method.
- Deduplicate: Check for duplicates against the existing transaction log. Flag any transaction that already appears.
- Auto-categorize each transaction using these rules:
- Software/subscriptions →
Software & Subscriptions - Payroll/contractors →
Payroll & Contractors - Rent/office →
Facilities - Marketing/ads →
Marketing - Travel/meals →
Travel & Entertainment - Legal/professional services →
Professional Services - Loan payments/interest →
Debt Service - Revenue from customers →
Revenue - Investments/grants →
Other Income - Everything else → assign best-guess category and flag for review.
- Software/subscriptions →
- Flag anomalies:
- Duplicate amounts to the same vendor in the same week.
- Transactions over an outlier threshold (3x the median of the same category).
- Round numbers that look like estimates rather than actuals.
- Missing counterparty or description.
- Prepare the output — a categorized transaction log.
- (Optional) Post to ledger — hand off to
ledger-management.
Output Format
A structured table (Markdown or CSV):
| Date | Description | Counterparty | Amount | Category | Confidence | Flag |
| 2026-04-01 | Stripe payout | Stripe | $12,450.00 | Revenue | HIGH | — |
| 2026-04-02 | AWS Invoice #8932 | AWS | $3,421.00 | Software & Subs | HIGH | — |
| 2026-04-03 | Lunch — Joe's Diner | Joe's Diner | $42.18 | Travel & Entertain | MEDIUM | REVIEW: personal vs business? |
Summary statistics at the top:
- Total transactions processed
- By category (pie or table)
- Flagged count / review-needed count
- Uncategorized count
Done Criteria
The skill is complete when:
- All transactions in the source data are parsed, deduplicated, and categorized
- Anomaly flags are applied and surfaced to the user
- Summary statistics are provided (total, by category, flagged count)
- Every transaction has a confidence score (HIGH, MEDIUM, or LOW)
- The categorized log is ready for hand-off to
ledger-management - No transaction is auto-posted to the ledger without human review when flags exist
Pitfalls
- Categorizing by payment method instead of nature: Categorizing everything from the corporate Amex as "Corporate Card" rather than splitting into actual expense categories (travel, software, meals)
- Forcing every transaction to HIGH confidence: Overriding the system's MEDIUM/LOW flags to avoid review items. Unreviewed medium-confidence items silently corrupt the general ledger
- Skipping deduplication on manual entries: Assuming manually-entered transactions are always unique. Manual entries are the most common source of duplicates — always scan
- Lumping bank fees into a single "Bank Charges" line: Payment processor fees, wire fees, FX markups, and account maintenance fees are fundamentally different costs and should be split if material
- Auto-posting without human review when flags exist: Processing a transaction batch with 3+ flagged items straight to the ledger without pausing for review. Flags exist for a reason — every flagged batch deserves a human sanity check before ledger posting
Verification
Can you trace any transaction from source data through categorization to the output table? Are all anomaly flags explained with a reason? Is the confidence score justified for each categorization? If a transaction is flagged MEDIUM or LOW, is there a clear path for the user to resolve it?
Example
Example 1: CSV Upload
User: "Here's our April bank export CSV — can you process and categorize all transactions?"
→ You parse the CSV, detect format (date/description/amount columns), deduplicate against prior months, auto-categorize each line item, flag 3 anomalies (one duplicate vendor payment, one $8,400 round-number transfer, one missing counterparty), and present a categorized table with summary stats and flags highlighted for review.
Example 2: Receipt Processing
User: "I have 12 receipts from our team's offsite — process these"
→ You extract dates, vendors, and amounts from the receipts, categorize all as Travel & Entertainment, apply the $500/receipt founder-scrutiny rule (none flagged since amounts are small), deduplicate against the period, and present a clean categorized batch ready for ledger posting.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

