From our LinkedIn post
Your team is giving AI access to company files. These six Claude Code mods make it ask before deleting, hide client details, track cost and log every action.
Claude Code is the version of Claude that works on your computer: it opens files, edits them and runs programs. Since 1 October 2026, mods let you set rules for what it can do and show you what it is doing. A mod is a small program that plugs into Claude Code.
lv-ask-before-deleteBefore Claude deletes or replaces files, it stops and shows exactly what would go, for example "4 files in 2 folders", and waits for a yes. No answer means no delete.
Also catches database wipes, replaced files and lost unsaved work.
lv-client-data-shieldEmails, phone numbers, card and bank numbers become labels like [email 3] and [card 1] before Claude reads a file. Claude still does the analysis without seeing the real details. Money amounts stay visible.
Limit: Names and addresses are not hidden.
lv-cost-meterA running cost against a budget you set, shown above the prompt. Over budget, Claude must ask before each step. Set the budget with /budget 10 (default $5).
Limit: On Pro and Max plans the figure is the equivalent API price, not money charged. It runs one step behind.
lv-work-receiptAfter every answer, a one-line receipt: files changed, commands run, web pages fetched and anything blocked. The full receipt is saved to a dated file in the project. Type /receipt to see the last one.
lv-team-commands/weekly-report, /client-brief and /meeting-actions send one company-standard prompt, so everyone gets the same format. The wording lives in one file your team can edit. No coding.
lv-audit-logEvery action Claude takes, written to a CSV file that opens in Excel: time, project, tool, what it did, and whether it was allowed, denied or failed.
Plus example settings for IT teams that limit which mods staff can install. It is an example; we have not used it to lock down a live team.
Built and tested by Lets Viz on Claude Code 2.1.288 (Windows 11), with fake client data in a spare folder and Anthropic's test kit (29 automated tests). Shared as is, with no warranty.
Small programs that plug into Claude Code and change how it behaves, for example asking before a delete or hiding client data. Anthropic added them on 1 October 2026 (v2.1.287).
No. They only work in Claude Code, in a terminal or the Code tab of the desktop app.
A mod is not sandboxed: it runs with your access to the computer. Only install mods you trust, and have someone technical read the code first.
Yes. An administrator can allow only approved mods through Claude Code's managed settings. The kit includes an example settings file.
A settings hook runs a command, web request or prompt when something happens in a session. A mod is a set of small functions inside a plugin that runs in Claude Code's own process, so it can also rewrite what Claude is about to do, draw new panels or replace a built-in feature. In Anthropic's words, hooks can't rewrite events, draw new UI or replace features; mods can.
For a business team, ask-before-deleting. It stops Claude before it deletes or replaces files and waits for a yes, which covers the mistake that costs the most. Add the others once the team is used to it.
Book a free 30-minute call. Tell us what your team is trying to do and we will tell you honestly what it takes.
Book a free call