claude-game-re-plugin

What You Get

This is the write guard. It was given the call Claude would make to set the health value in the tutorial that ships with Pointer Lab. This is real output. Exit code 2 blocks the call, and Claude reads the message.

$ echo '{"hook_event_name":"PreToolUse","tool_name":"mcp__pointerlab__write","tool_input":{"address":"pointerlabtutorial.exe+2DBC8","type":"f32","value":"1000"}}' | powershell.exe -NoLogo -NoProfile -NonInteractive -ExecutionPolicy Bypass -File hooks/guard-writes.ps1; echo "exit $?"
game-re blocked mcp__pointerlab__write: it writes a value into the target's memory.

Pointer Lab's write, patch, allocate, inject and thread tools are off until the user opts in. This is the game-re plugin's safety default, not a Pointer Lab error. Do not retry the call or look for another way to make the change. Tell the user what was blocked and how to opt in.

To opt in, and only for software they own or the Pointer Lab tutorial, the user closes Claude Code and starts it again with GRC_ALLOW_WRITES set to 1:
  PowerShell:  $env:GRC_ALLOW_WRITES = "1"; claude
  cmd:         set "GRC_ALLOW_WRITES=1" then claude
  Git Bash:    GRC_ALLOW_WRITES=1 claude

Reads, scans, pointer scans, the access watch and hardware breakpoints work without it. Undo is never blocked: patch_restore_all, patch_remove, set_frozen with frozen false, and aa_disable_all.
exit 2

Claude reads the message and relays it to you. It tells Claude not to retry the call or look for another way to make the change. It tells you how to opt in, and what still works without it.

How It Works

Item Specifics

Before You Download