agentleFS
bkjain655bkjain655/agent-skills/skills/git-repo-privacy-audit/SKILL.md

git-repo-privacy-audit

Skill0 starsChanged 4 months agoMITView on GitHub

Audit a git repository for secrets, API keys, and personal information in both the working tree AND the full commit history (including commit author/committer metadata), then safely scrub anything found and prevent recurrence. Use before making a repo public, after a suspected leak, or when someone asks whether a repo exposes secrets or personal data. Framework-agnostic.

  • Reads credentials
  • Deletes or force-pushes
  • Commits and pushes

The file

We split this skill into the rules it gives an agent

1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Distinctive:

    Run Part 1 [a]–[e]. If a real secret/key is or ever was committed: rotate it first (scrubbing history does not un-leak an exposed …

    line 113

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Git Repo Privacy Audit & Scrub

Two separate questions, often conflated: (1) secrets/keys (credentials — high severity) and (2) PII (emails, phone, names — moderate). Both can hide in file content OR in commit metadata. Check all four combinations.

Part 1 — Audit (read-only)

# [a] Is a real .env (not .env.example) tracked?
git ls-files | grep -E '(^|/)\.env' | grep -v '\.env\.example'   # want: empty

# [b] Secret patterns across ALL history (added or later removed), placeholders excluded
PATS='re_[A-Za-z0-9]{16,}|AIza[0-9A-Za-z_-]{20,}|sk_(live|test)_[A-Za-z0-9]{10,}|AKIA[0-9A-Z]{16}|ghp_[A-Za-z0-9]{20,}|github_pat_|xox[baprs]-|-----BEGIN [A-Z ]*PRIVATE KEY'
git log --all -p | grep -aInE "$PATS" | grep -aivE 'xxxx|placeholder|example|your[-_]'

# [c] PII in file content — use FIXED strings and a reliable per-blob check.
#     grep regex '.' matches any char and inflates counts; -F avoids that.
git log --all --oneline -S"person@gmail.com" -- path/to/file   # pickaxe = file-content changes only

# [d] PII in COMMIT METADATA (the one everyone forgets — GitHub shows this publicly):
git log --all --format='author: %an <%ae>' | sort | uniq -c
git log --all --format='commit: %cn <%ce>' | sort | uniq -c

# [e] .env.example must contain only placeholders
git show HEAD:.env.example

Key insight: a repo's code can be clean while its commit authorship leaks a personal email — because git log/GitHub display author.email. Always run [d].

Measurement traps that cause false positives/negatives:

  • grep 'a.b@x.com' — . is any-char; use grep -F.
  • git log -p | grep counts diff context lines and author lines, so a string present only in metadata shows up as many "hits". Cross-check with pickaxe -S (file content only) and per-blob git show <sha>:<path> | grep -F.
  • zsh: for c in $(git rev-list --all) splits oddly — prefer git rev-list --all | while IFS= read -r c.
  • Definitive check: git show <sha>:<path> > /tmp/f && grep -Fn "STRING" /tmp/f.

Part 2 — Scrub (rewrites history → force-push)

git filter-repo is best if installed. If not, git filter-branch works.

Scrub PII from FILE content (replace, don't delete lines — keeps old commits valid). Use perl, not sed -i '' — GNU sed on the PATH mis-parses the empty suffix and silently does nothing. In perl regex, escape @ as \@ (arrays interpolate even in single-quoted -e).

# scrub.sh  (referenced by absolute path so cwd doesn't matter)
if [ -f path/to/file ]; then
  perl -i -pe 's/old\.value\@gmail\.com/redacted\@example.com/g' path/to/file
fi
exit 0
FILTER_BRANCH_SQUELCH_WARNING=1 git filter-branch -f --tree-filter '/abs/scrub.sh' -- --all

Scrub a personal email from AUTHOR/COMMITTER metadata → replace with GitHub's privacy noreply so commits still link to the account (get the numeric id from gh api users/<username> --jq .id → <id>+<username>@users.noreply.github.com):

FILTER_BRANCH_SQUELCH_WARNING=1 git filter-branch -f --env-filter '
if [ "$GIT_AUTHOR_EMAIL" = "old@gmail.com" ]; then export GIT_AUTHOR_EMAIL="ID+user@users.noreply.github.com"; fi
if [ "$GIT_COMMITTER_EMAIL" = "old@gmail.com" ]; then export GIT_COMMITTER_EMAIL="ID+user@users.noreply.github.com"; fi
' -- --all

Always finish the cleanup — filter-branch leaves the old commits reachable via backup + remote-tracking refs, so scans keep "finding" them until you purge:

git for-each-ref --format='%(refname)' refs/original/ refs/remotes/ | xargs -r -n1 git update-ref -d
git reflog expire --expire=now --all
git gc --prune=now

Part 3 — Verify, then push

git log --all -p | grep -Fc "STRING"                 # 0
git log --all --format='%ae %ce' | sort -u           # only the safe address(es)
# push the rewritten history; --force-with-lease aborts if the remote moved:
REMOTE=$(git ls-remote <url> refs/heads/main | cut -f1)
git push --force-with-lease=main:$REMOTE <url> main
# confirm on the PUBLIC side (not just local):
gh api repos/<owner>/<repo>/commits --jq '.[].commit.author.email' | sort -u

Part 4 — Prevent recurrence

The personal email in commits comes from local git config. Set it to the noreply:

git config --global user.email "ID+username@users.noreply.github.com"
# GitHub: Settings → Emails → "Keep my email addresses private" to enforce it.

Before making any repo public — the gate

Run Part 1 [a]–[e]. If a real secret/key is or ever was committed: rotate it first (scrubbing history does not un-leak an exposed credential — assume it's compromised). Then scrub, verify, and only then flip visibility. Making a repo public is effectively irreversible (clones/caches during the public window persist).

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.