Small git tools to take authorship from LLM work

A developer created a Git tool called git_reauthor.sh that rewrites commit authorship and Co-Authored-By trailers for a range of commits. The script can replace author/committer identity, strip or replace Co-Authored-By lines, and is intended to help developers take authorship credit for work originally generated by large language models.

| /usr/bin/env bash | | | | git reauthor.sh <back until sha inclusive <committer name email <coauthored by name email | | | | Rewrite the identity author AND committer name/email of every commit from | | <back until sha inclusive up to HEAD, and fix up the Co-Authored-By trailer. | | | | Arguments 2 and 3 are optional: | | no <committer name email - use the current git config user.name/email | | as the identity and, since arg 3 is then also absent, strip co-authors . | | no <coauthored by name email - strip every Co-Authored-By line from the | | commit messages. | | <coauthored by name email present - replace any existing Co-Authored-By | | line s with a single "Co-Authored-By: <that value " appended if the | | commit had none . | | | | Identity arguments are in the usual git form: Name Surname <email@host | | Commit dates are left untouched. | | | | WARNING: this rewrites history commit SHAs change . filter-branch keeps a | | backup at refs/original/... ; undo with: | | git reset --hard refs/original/refs/heads/<branch | | | | set -euo pipefail | | | | if "$ " -lt 1 || "$ " -gt 3 ; then | | echo "usage: $ basename "$0" <back until sha inclusive <committer name email <coauthored by name email " &2 | | echo " args 2 and 3 optional; omit arg 3 to strip Co-Authored-By lines," &2 | | echo " omit arg 2 to also fall back to the current git config identity " &2 | | exit 2 | | fi | | | | SINCE="$1" | | IDENT="${2:-}" | | COAUTHOR="${3:-}" | | | | --- validate ------------------------------------------------------------- | | if git rev-parse --is-inside-work-tree /dev/null 2 &1; then | | echo "error: not inside a git work tree" &2 | | exit 1 | | fi | | | | if git rev-parse --verify --quiet "${SINCE}^{commit}" /dev/null; then | | echo "error: '$SINCE' is not a valid commit" &2 | | exit 1 | | fi | | | | Resolve the target identity: an explicit "Name <email " argument, or the | | current git config user.name/user.email when arg 2 is omitted. | | if -n "$IDENT" ; then | | case "$IDENT" in | | "<" " " : ;; | | echo "error: committer identity must look like 'Name <email ', got '$IDENT'" &2; exit 1 ;; | | esac | | RA NAME="$ printf '%s' "$IDENT" | sed -E 's/ :space: < ^ :space: $//' " | | RA EMAIL="$ printf '%s' "$IDENT" | sed -E 's/^ ^< < ^ . $/\1/' " | | else | | RA NAME="$ git config user.name || true " | | RA EMAIL="$ git config user.email || true " | | if -z "$RA NAME" || -z "$RA EMAIL" ; then | | echo "error: no identity given and git config user.name/user.email is unset" &2 | | exit 1 | | fi | | fi | | | | if -n "$COAUTHOR" ; then | | case "$COAUTHOR" in | | "<" " " : ;; | | echo "error: co-author must look like 'Name <email ', got '$COAUTHOR'" &2; exit 1 ;; | | esac | | fi | | | | Range that includes SINCE itself. If SINCE is the root commit it has no | | parent, so fall back to the full history reachable from HEAD. | | if git rev-parse --verify --quiet "${SINCE}^{commit}^" /dev/null; then | | RANGE="${SINCE}~1..HEAD" | | else | | RANGE="HEAD" | | fi | | | | --- awk program that rewrites the commit message ------------------------- | | Removes existing Co-Authored-By lines, trims trailing blank lines, and when | | ca is non-empty appends a single Co-Authored-By trailer, keeping it inside | | the trailing trailer block when one already exists. | | AWK PROG="$ mktemp "${TMPDIR:-/tmp}/reauthor.XXXXXX" " | | trap 'rm -f "$AWK PROG"' EXIT | | cat "$AWK PROG" <<'AWK' | | { buf NR = $0 } | | END { | | m = 0 | | for i = 1; i <= NR; i++ { | | if tolower buf i ~ /^co-authored-by:/ continue | | out ++m = buf i | | } | | while m 0 && out m ~ /^ \t $/ m-- trim trailing blanks | | for i = 1; i <= m; i++ print out i | | if ca = "" { | | if m 0 && out m ~ /^ A-Za-z0-9- +: \t / { last line already a trailer | | print "Co-Authored-By: " ca | | } else { | | if m 0 print "" blank line before trailer | | print "Co-Authored-By: " ca | | } | | } | | } | | AWK | | | | --- rewrite -------------------------------------------------------------- | | export RA NAME RA EMAIL COAUTHOR | | | | ENV FILTER=' | | export GIT AUTHOR NAME="$RA NAME" | | export GIT AUTHOR EMAIL="$RA EMAIL" | | export GIT COMMITTER NAME="$RA NAME" | | export GIT COMMITTER EMAIL="$RA EMAIL" | | ' | | MSG FILTER="awk -v ca=\"\$COAUTHOR\" -f \"$AWK PROG\"" | | | | if -n "$COAUTHOR" ; then | | echo "Reauthoring $RANGE as '$RA NAME <$RA EMAIL ', co-author - '$COAUTHOR'" | | else | | echo "Reauthoring $RANGE as '$RA NAME <$RA EMAIL ', stripping Co-Authored-By lines" | | fi | | | | FILTER BRANCH SQUELCH WARNING=1 git filter-branch -f \ | | --env-filter "$ENV FILTER" \ | | --msg-filter "$MSG FILTER" \ | | -- $RANGE | | | | echo | | echo "Done. Backup ref kept at refs/original/. Undo with:" | | echo " git reset --hard refs/original/refs/heads/$ git rev-parse --abbrev-ref HEAD " |