Nine skills, free to download, that run your firm's workflows on top of Bizora's research.
Last week we showed a Claude skill reviewing a Form 1040 against the preparer's workpapers. The most common reply we got was some version of the same question: can you just send us the skill?
So we packaged them. The Bizora Skills Library is live today with nine skills, all free, and we are adding to it based on what users ask for.
A skill is a set of instructions you upload once into Claude that tells it how to run a specific piece of work. Not a prompt you retype every time, and not a chatbot you have to steer turn by turn. It carries the whole procedure: what to ask for up front, what order to check things in, what to flag, what to escalate, and what the finished deliverable should look like.
Once a skill is loaded, you drop in the return, the notice, the K1s, or whatever the work starts with, and it runs.
The skills in this library are built to call Bizora underneath. When one of them hits a question that needs authority, the research runs against our tax database and the citation that comes back points at the actual source, the same one you land on when you click a citation in the web app. The model is not recalling what it thinks the rule says.
When we started building these internally, we assumed the research would be the hard part. It was not. Bizora already handles that.
The hard part was the ordering.
Take a K1 tie out. Tracing the amounts into the return is mechanical. Knowing that basis comes before at risk, and at risk before the passive limitation, and that a loss suspended at one gate never reaches the next, is the part a second year associate gets wrong and a manager catches on review. Same with a notice response, where the first thing that matters is not the substance of the IRS position but how many days are left on the response window and what right expires when the clock runs out.
That sequencing usually lives in one partner's head and nowhere else in the firm. Writing it into a skill is what makes the work repeatable for everyone else.
Form 1040 Reviewer. Reviews a prepared return against the preparer's workpapers, ties every material line back to its support, and returns an exception report graded by severity plus a printable review dashboard for the engagement file.
Multiyear Return Comparison. Compares two or more years of any return type and flags every material movement, then tells you whether the file explains it, a rule change explains it, or nothing does. Catches broken carryforwards, items that quietly disappeared, and elections a new preparer never knew about.
K1 Tie Out. Traces each passthrough K1 into the individual return code by code, confirms every amount landed on the correct form, runs the basis, at risk and passive limitation gates in order, and writes the basis questions to send back to the client.
IRS Notice Response. Decodes the notice, leads with what right expires and when, ties each proposed change back to the filed return, recomputes the proposed tax rather than accepting the IRS figure, and drafts a complete response package for your signature.
Tax Memo Builder. Turns a question into a finished, cited memo in your firm's format, with a committed fact pattern, contrary authority addressed inside the analysis, and a stated confidence level. Handles memos to the file, to the client, and to another professional.
Authority Level Assessment. Grades a position against the confidence standards that drive penalty exposure and disclosure, weighing supporting and contrary authority side by side rather than counting citations, and separating taxpayer exposure from preparer exposure.
Entity Structure Analysis. Models the alternatives side by side including payroll and self employment tax, the QBI interaction, state entity level taxes and passthrough entity elections, conversion cost, break even and exit. Reasonable compensation is modeled across a defensible range instead of one chosen number.
Tax Scenario Comparison. Builds the client facing before and after: a baseline plus one tab per planning strategy, each with its own tax total, the difference against the baseline, a plain language takeaway and a citation. Strategies that raise this year's tax get labeled as costs rather than savings.
Client Data Request. Builds a client specific document request and questionnaire off the prior year return instead of a generic organizer, splitting what the client has to go find from what only the client knows, and marking what actually blocks filing.
There is a short walkthrough video at the top of the library page showing the upload step if this is your first time.
The skills are written for Claude, but the logic inside them is portable. Open one, copy the instructions, and paste them into ChatGPT or whatever assistant your firm has standardized on. Anywhere you can upload a skill or set up a project with saved instructions, these will run.
The one thing that has to stay connected is Bizora. The research and the citations come from the Bizora API and MCP, which is what keeps the output defensible no matter which assistant sits on top of it.
If you want the skill to run automatically on everything you send it, ask Claude to convert it into project instructions. Save that to a project and every file you drop in runs through the workflow without you loading the skill again.
Nine is where the library starts, not where it ends. Each of these came out of a real request from a user, and turning a described workflow into a working skill takes us very little time.
If there is something your team runs every week by hand, describe it to us and we will build it. Email adam@bizora.ai or reply to the weekly newsletter.
Download the Skills Library or start a free trial if you are not on Bizora yet.
Bizora AI turns hours of manual research into seconds, with every answer backed by primary source citations. Start your 7-day free trial. No credit card required.
Start Free TrialRelated articles