This is a request about fair attribution, not a new vulnerability report.
On 2026-01-27 I reported that executable-name-only validation in ALLOW_COMMANDS does not constrain execution, because allowlisted binaries expose argument-level exec primitives (GHSA-jm26-853c-62h9, unpublished). The report used git -c alias.x=!... and named find -exec, awk system() and python -c as the same class. It was closed the following day as expected behaviour, on the basis that ALLOW_COMMANDS was not a security boundary.
Since then:
The project now documents argument hardening in SECURITY.md, implements it in command_validator.py for git, find, awk and tar (#40/#41), and has published two High severity advisories for gaps in it.
Given that the class is indeed in scope, would you consider publishing GHSA-jm26-853c-62h9 as a historical advisory against 1.0.3, or adding an acknowledgment line to SECURITY.md? I raised this on the advisory thread but it may not have surfaced given the advisory is closed. Either is fine. Thanks for considering it.
This is a request about fair attribution, not a new vulnerability report.
On 2026-01-27 I reported that executable-name-only validation in ALLOW_COMMANDS does not constrain execution, because allowlisted binaries expose argument-level exec primitives (GHSA-jm26-853c-62h9, unpublished). The report used
git -c alias.x=!...and namedfind -exec,awk system()andpython -cas the same class. It was closed the following day as expected behaviour, on the basis that ALLOW_COMMANDS was not a security boundary.Since then:
gitcannot be used to run arbitrary commands, usinggit -c alias.pwn=!sh -c "touch marker" pwnas the example.gitconfig flags,find -exec/-execdir,awksystem(), andtarexec options.The project now documents argument hardening in SECURITY.md, implements it in
command_validator.pyforgit,find,awkandtar(#40/#41), and has published two High severity advisories for gaps in it.Given that the class is indeed in scope, would you consider publishing GHSA-jm26-853c-62h9 as a historical advisory against 1.0.3, or adding an acknowledgment line to SECURITY.md? I raised this on the advisory thread but it may not have surfaced given the advisory is closed. Either is fine. Thanks for considering it.