Format string vulnerability in the helptags_one function in src/ex_cmds.c in Vim 6.4 and earlier, and 7.x up to 7.1, allows user-assisted remote attackers to execute arbitrary code via format string specifiers in a help-tags tag in a help file, related to the helptags command.

The sandbox for vim allows dangerous functions such as (1) writefile, (2) feedkeys, and (3) system, which might allow user-assisted attackers to execute shell commands and write files via modelines.

vim 6.3 before 6.3.082, with modelines enabled, allows external user-assisted attackers to execute arbitrary commands via shell metacharacters in the (1) glob or (2) expand commands of a foldexpr expression for calculating fold levels.

The (1) tcltags or (2) vimspell.sh scripts in vim 6.3 allow local users to overwrite or create arbitrary files via a symlink attack on temporary files.

VIM before 6.3 and gVim before 6.3 allow local users to execute arbitrary commands via a file containing a crafted modeline that is executed when the file is viewed using options such as (1) termcap, (2) printdevice, (3) titleold, (4) filetype, (5) syntax, (6) backupext, (7) keymap, (8) patchmode, or (9) langmenu.

vim 6.0 and 6.1, and possibly other versions, allows attackers to execute arbitrary commands using the libcall feature in modelines, which are not sandboxed but may be executed when vim is used to edit a malicious file, as demonstrated using mutt.

vim (aka gvim) allows local users to modify files being edited by other users via a symlink attack on the backup and swap files, when the victim is editing the file in a world writable directory.

vim (aka gvim) processes VIM control codes that are embedded in a file, which could allow attackers to execute arbitrary commands when another user opens a file containing malicious VIM control codes.