Skills library · built for Claude
AI Skills built for internal audit, SOX, and IT controls testing.
How these hold up under review
Every workpaper has to survive someone senior asking where a number came from — and, in time, a regulator or quality reviewer asking the same. Three things make that answerable here.
A written record of what changed
A Skill is a text document. When your testing steps change, that shows up as a change you can read — which rule, and when. Keep it wherever your firm already keeps its methodology documents.
The fixed parts are computed, not reasoned
Sampling, date handling, totals, threshold tests — anything with one correct answer is calculated. Same inputs, same output, every run. AI judgment is reserved for where a fixed answer is not possible.
Every item points back to its source
Skills are written to cite the line, section, or sample ID behind each finding, so a reviewer checks the source rather than taking the draft at its word.
What comes out is an audit input — organized, traceable, and ready to review. The auditor reaches every conclusion and signs everything.
Shared building blocks
The core logic every skill runs on — written once, tested independently, and reused rather than rebuilt. This is where the reproducibility actually comes from.
These aren't downloaded separately — they come bundled inside every skill that uses them.
Sampling Engine
BuiltPicks statistically sound samples using the same accepted audit sampling methods every time (dollar-weighted, attribute, random), so results can be reproduced and defended. Fully tested — 30 of 30 checks passing.
30 / 30 checks passing
Time Zone & Date Safety Check
BuiltConverts every timestamp to one common time zone before comparing dates, so a legitimate entry in a different time zone never gets wrongly flagged — or a real problem missed. Proven across three different skills, including catching a live mismatch between a company's stated time zone and its real address.
Repeat-Item Matcher
BuiltMatches this year's items to last year's automatically — exact matches first, close-but-not-exact matches flagged for a person to confirm, never assumed. Fully tested — 11 of 11 checks passing.
11 / 11 checks passing
Cross-System Identity Matcher
BuiltMatches the same person across two systems that use different ID numbers or formats. Any uncertain match goes to a person to accept or reject before it's used. Fully tested — 19 of 19 checks passing.
19 / 19 checks passing
Built-In Verification Standard
BuiltA quality bar every workpaper has to clear before it ships: results are checked against a genuinely independent source, and every automated check is proven able to catch a real mistake, not just designed to look like it works. Two new skills are now built to this standard from day one.
Test Plan Generator
BuiltTurns a plain-English control description into a documented test plan, which a person then has to approve before any testing begins.
Excel Workpaper Builder
In progressBuilds the finished Excel workpaper — header, tick marks, sign-off line — in Auditshore's standard format. Currently rebuilt individually inside six different skills; next up is turning it into one shared tool every skill can use.
Deficiency Severity Calculator
BuiltClassifies a control gap as a deficiency, significant deficiency, or material weakness from how much money it put at risk and the risk rating already recorded, against thresholds locked and fingerprinted before the run. Ships inside SOX Deficiency Evaluation.
Want one of these fitted to your methodology?
Every skill assumes a house format somewhere — your control matrix, your workpaper template, your thresholds. Adapting that is the work we do.