Automated Reconciliation

Match bank statements to the general ledger at scale: deterministic matching first, fuzzy matching with confidence scores second, and a clean exceptions list for human sign-off.

⚠️ **Finance warning / Avertissement finance** : informational only, not investment advice. A reconciliation that forces matches hides real errors and real fraud: a wrong match rate taken at face value can conceal missing cash, there is a real risk of loss in any close decision taken from an unreconciled balance, and no output here is a promise of return. Curated by Skill Harbor: the automated reconciliation skill of GAJETOso/financeskills. Use it to match disparate financial data sources at scale, bank statements against the general ledger or an ERP export, invoices against payments, without weeks of manual ticking. The skill is honest about its own technical limit up front: language models are not good at matching 50,000 rows, so large datasets run through Python libraries such as pandas and RecordLinkage, and the model is used where it is actually strong, resolving the ambiguous matches, roughly the 5 percent the code cannot settle. The framework runs in priority order: data cleaning first (standardizing vendor names, so AWS and Amazon Web Svcs become one counterparty), deterministic matching on exact identifiers, amounts and dates, probabilistic fuzzy matching on similar names with the same amount inside a two day window using Jaro-Winkler or Levenshtein distance, and exception handling that flags what could not be matched instead of forcing it. The technical steps add a tolerance window on amounts (within five cents for rounding), many-to-one resolution (one bank deposit that represents three separate ledger invoices, handled by the subset-sum engine in the shipped calculate.py script), and journal entry suggestions for bank fees or interest found in the statement but missing from the ledger. The deliverable is a reconciliation report: the match rate and total reconciled value, the unmatched items from both sources, the ambiguous matches requiring human sign-off with their confidence scores, and the suggested entries, with a template bank reconciliation statement and best practices for keeping the trail reviewable. From the GAJETOso/financeskills repository (MIT). Honest caveats: reconciliations support the cash balance in official financial statements, so have the exception handling and the final statement checked by a qualified professional before any official use; the statements and ledger exports are yours to supply complete and unaltered, and no ambiguous match should ever be posted without the human sign-off the report is built around. Skill Harbor never reviews the code, review it yourself before use.

At a glance

What: Match bank statements to the general ledger at scale: deterministic matching first, fuzzy matching with confidence scores second, and a clean exceptions list for human sign-off.
Cost: Free
Needs: Use "Automated Reconciliation" with your Muse.
Install: Copy the installer prompt below into your Muse — your agent does the rest.

⚠️ **Finance warning / Avertissement finance** : informational only, not investment advice. A reconciliation that forces matches hides real errors and real fraud: a wrong match rate taken at face value can conceal missing cash, there is a real risk of loss in any close decision taken from an unreconciled balance, and no output here is a promise of return. Curated by Skill Harbor: the automated reconciliation skill of GAJETOso/financeskills. Use it to match disparate financial data sources at scale, bank statements against the general ledger or an ERP export, invoices against payments, without weeks of manual ticking. The skill is honest about its own technical limit up front: language models are not good at matching 50,000 rows, so large datasets run through Python libraries such as pandas and RecordLinkage, and the model is used where it is actually strong, resolving the ambiguous matches, roughly the 5 percent the code cannot settle. The framework runs in priority order: data cleaning first (standardizing vendor names, so AWS and Amazon Web Svcs become one counterparty), deterministic matching on exact identifiers, amounts and dates, probabilistic fuzzy matching on similar names with the same amount inside a two day window using Jaro-Winkler or Levenshtein distance, and exception handling that flags what could not be matched instead of forcing it. The technical steps add a tolerance window on amounts (within five cents for rounding), many-to-one resolution (one bank deposit that represents three separate ledger invoices, handled by the subset-sum engine in the shipped calculate.py script), and journal entry suggestions for bank fees or interest found in the statement but missing from the ledger. The deliverable is a reconciliation report: the match rate and total reconciled value, the unmatched items from both sources, the ambiguous matches requiring human sign-off with their confidence scores, and the suggested entries, with a template bank reconciliation statement and best practices for keeping the trail reviewable. From the GAJETOso/financeskills repository (MIT). Honest caveats: reconciliations support the cash balance in official financial statements, so have the exception handling and the final statement checked by a qualified professional before any official use; the statements and ledger exports are yours to supply complete and unaltered, and no ambiguous match should ever be posted without the human sign-off the report is built around. Skill Harbor never reviews the code, review it yourself before use.

Version:

@

Created by: @gajetoso

Not verified? How to check before you trust →

Install — Free · open source (MIT) View source code

⌁

Install

Copy the install package below, then paste it into Muse

Wanna be extra careful?

The install prompt below already includes the vetting steps: your agent follows the community checklist before installing anything with executable code. Want more?

How to check a build before installing →

After installation, ask me if I want to leave private feedback
Private for now: only the Skill Harbor team sees it. Sellers may be looped in later.

Use "Automated Reconciliation" with your Muse. Prerequisites: both data sources in full, the bank statement as PDF or CSV and the general ledger or ERP export for the same period, plus your matching rules (which identifiers exist, the amount tolerance, the date window). Python with pandas is needed for large volumes: the deterministic and fuzzy engines do the bulk matching, and this skill produces the report structure and resolves the ambiguous tail. No bank login and no credentials are involved, exports only. Informational only, not investment advice, and have the output checked by a qualified professional before any official use. 1. Open the skill: https://github.com/GAJETOso/financeskills/blob/main/skills/automated-reconciliation/SKILL.md and copy the full SKILL.md text. 2. Paste it into a chat with Muse and add: "Reconcile these two sources: standardize the vendor names, run the deterministic matches on identifiers, amounts and dates, run the fuzzy pass inside the tolerance and date window, resolve the many-to-one deposits, list every unmatched item from both sides, score the ambiguous matches for my sign-off, and suggest the missing fee and interest entries." 3. Read the exceptions before the match rate: a high rate with a long exception list is a worse reconciliation than a lower rate fully explained. Tip: never let an ambiguous match post itself; the confidence score is a queue for your judgment, not an approval. Safety: a skill is plain-text instructions; it runs nothing by itself. Analysis and documents only, no orders and no account access. Informational only, not investment advice.

?

Questions

How do I install a build?

Every product page includes a copy-paste install prompt. Paste it into your Muse and it sets the build up for you — no manual configuration.

Where does my money go?

Straight to the seller. Skill Harbor never processes payments: checkout happens on the seller’s own page, usually Stripe.

What does the ✓ next to a creator’s name mean?

It means we confirmed the identity of the person behind the listing. It says nothing about the code itself — always check a build before installing it.