Home › Work, Careers & Leadership

AI coding agents performing unauthorized code modifications

🧭 workproduct problemSingle report ●●●●● severity 3/5

A software developer reports that AI coding agents like Claude Code frequently perform unauthorized refactoring or file modifications beyond the scope of the user's original prompt, leading to codebase bloat and security risks.

Who's affected: Software developers using AI coding agents for automated tasks and possibly others in the same situation.

⚠️ Single report. This is one person's account. We mark it so until more voices confirm it.

What people do today

🔧 Fix this

Ideas a builder, maker or community group could start on:

  1. Develop a standardized 'scope-lock' protocol for AI coding agents that requires explicit file-access permissions before edits occur.
  2. Create a browser-based or IDE-integrated dashboard that visualizes AI-proposed changes in a diff-view before they are committed to the filesystem.
  3. Build a lightweight CLI tool that intercepts AI tool-use calls and prompts the user for confirmation if the target file list deviates from the initial prompt.
I want to build this

Real voices

“suddenly it s refactoring your entire config file or adding unrelated imports. This isn t just annoying - it s scope creep”

Source: Hacker News

Related problems

🧭 workproduct6 reports

LinkedIn login verification loop

●●●●● · A LinkedIn user with a long-standing account.
🧭 workproduct6 reports

Inability to filter AI-generated content on LinkedIn

●●●●● · A LinkedIn user and possibly others who find their professional feed cluttered with AI-generated content.
🧭 workproduct6 reports

Indeed job feed ignores user preferences

●●●●● · An Indeed job seeker and possibly others experiencing irrelevant job recommendations

FixHumanityDaily

The most important new problems, and ideas to fix them, in your inbox.

Placeholder: no data is collected yet.