GHSA-fx3j-rwgx-fr94: GitPython: separate_git_dir leaks env vars via crafted .gitmodules submodule name
GitPython up to 3.1.62 expands environment variables in the separate_git_dir path used by Repo._clone(). A malicious repository can name a submodule `$SOME_VAR`, and a normal clone plus submodule update then turns the victim's variable value into a directory name and, at DEBUG logging, into a log line. Upgrade to GitPython 3.2.0.
Published Oct 27, 2026 · Analysis Nov 1, 2026
What an attacker can do
Repo._clone() passes separate_git_dir through Git.polish_url() without expand_vars=False. The sibling fixes for GHSA-94p4-4cq8-9g67 and GHSA-rwj8-pgh3-r573 added that flag to the URL call sites, but not this one. The value comes from Submodule._clone_repo(), which builds it from the raw submodule name in .gitmodules. The name check added for GHSA-hmq2-w58f-27jc rejects absolute paths, drive letters and .., but not $.
The attacker only needs the victim process to clone their repository and run submodule_update(init=True, recursive=True). The reporter used git submodule add --name '$GPY_SECRET_MARKER'. The victim's variable value then appeared as a directory under .git/modules/, and it persists as long as the clone does. With the git.cmd or git.repo.base loggers at DEBUG, the Popen(...) line also prints the expanded --separate-git-dir= argument, because only URL passwords and Authorization headers are redacted.
The reporter notes the call then fails with an OSError, so the leak is a side effect. It matters most in CI jobs where tokens such as cloud keys sit in the environment and the attacker can read the resulting directory names or logs.
Who is affected
GitPython 3.1.62 and earlier, on any platform where git is 1.7.5 or newer (the default gitfile submodule path). Services and CI scripts that clone untrusted repositories and update submodules are the exposed pattern. Check the installed version:
pip show GitPythonGitPython
- Ecosystem
- PyPI
- Affected
- <= 3.1.62
- Fixed in
- 3.2.0
- Check
pip show GitPython
How to fix it
Upgrade to GitPython 3.2.0. The fix is pull request 2255, which keeps the separate git directory path literal.
pip install --upgrade "GitPython>=3.2.0"The 3.2.0 release also drops Python 3.7 and changes config parsing, so test before rolling out. The sources list no workaround. Until you upgrade, avoid submodule updates on untrusted repositories and keep secrets out of the environment of that process.
How to tell if you were hit
List .git/modules/ in clones made from third-party repositories. A directory named like a secret value, or one that does not match any submodule name in .gitmodules, points to this. Check .gitmodules for section names containing $. In CI logs with DEBUG enabled, search for Popen( lines holding --separate-git-dir= and see whether the path contains values from your environment. Rotate any secret that appears.
Original advisory
From the GitHub Advisory Database (CC BY 4.0).
Summary
Repo._clone()'s handling of separate_git_dir never disables Git.polish_url()'s environment-variable expansion, unlike every other caller of that helper that handles attacker-influenced data. Since separate_git_dir can be derived from an attacker-controlled .gitmodules submodule name, a completely ordinary clone + submodule_update(init=True, recursive=True) on a malicious repository expands an environment variable of the attacker's choosing and exposes its value both as a persistent on-disk directory name and, whenever DEBUG-level logging is enabled (a common CI/verbose-logging configuration), in a Popen(...) log line that is not redacted the way credential-bearing arguments are.
Root cause / Details
git/repo/base.py's _clone() has:
sep_dir = kwargs.get("separate_git_dir")
if sep_dir:
kwargs["separate_git_dir"] = Git.polish_url(sep_dir)This is the only remaining Git.polish_url() call site handling attacker-influenced data that does not pass expand_vars=False (there is one other unguarded call site, clone_path at line 1537, but its input is the caller's own local destination path, not attacker-controlled). The sibling fixes for GHSA-94p4-4cq8-9g67/GHSA-rwj8-pgh3-r573 added expand_vars=False to clone_url (5 lines below this one, in the same function), git/repo/base.py:1599, remote.py:814, and submodule/base.py:686 — this one did not receive it. (submodule/root.py also currently sets expand_vars=False at a related call site, but that was added incidentally by a later, unrelated commit, not as part of these security fixes.) Git.polish_url()'s docstring itself says: "Disable this for values obtained from an untrusted source, such as remote URLs." Its non-Cygwin branch does plain os.path.expandvars() on the entire string, including embedded variables anywhere in an absolute path.
separate_git_dir reaches _clone() from Submodule._clone_repo() (git/objects/submodule/base.py) as module_abspath = osp.join(parent_repo.git_dir, "modules", name), where name is the raw .gitmodules submodule section name — fully attacker-controlled content of a cloned repository. _validated_name() (added by the fix for GHSA-hmq2-w58f-27jc) only rejects names that are absolute, contain a drive letter, or contain .. path components; it does not reject $. _need_gitfile_submodules() is true for git ≥ 1.7.5, i.e. the default path for virtually every real-world installation.
Reproduction (live, against real, unmodified GitPython code, no monkeypatching)
Built a hostile repository with a plain, ordinary git command: git submodule add --name '$GPY_SECRET_MARKER' ../evil_submodule sub. Set an environment variable in the "victim" process (standing in for a real CI secret such as AWS_SECRET_ACCESS_KEY): GPY_SECRET_MARKER=AKIA_TOTALLY_FAKE_BUT_SECRET_LOOKING_VALUE_12345. Then ran the exact ordinary workflow the GHSA-hmq2-w58f-27jc advisory's own PoC also uses:
repo = Repo.clone_from(evil_parent_url, victim_clone_dir)
repo.submodule_update(init=True, recursive=True)Observed result (ls .git/modules/):
.git/modules/AKIA_TOTALLY_FAKE_BUT_SECRET_LOOKING_VALUE_12345/ (a full git repo: HEAD, config, objects/, refs/, hooks/)The literal $GPY_SECRET_MARKER-named directory was never created — only the expanded one exists, proving expansion happened before the value reached the git clone --separate-git-dir=... subprocess argument. The persistent, expanded-name directory under .git/modules/ is itself the primary disclosure channel — it survives on disk for as long as the clone does. A second, independent channel: with the git.cmd/git.repo.base loggers at DEBUG level (a common CI/verbose-logging setting, not requiring any special flag), GitPython's own Popen(...) debug log line prints the fully expanded --separate-git-dir=<value> subprocess argument verbatim — remove_password_if_present() only redacts URL passwords/Authorization headers, not path-style arguments like this one, so the expanded value is not scrubbed. (The call also fails downstream with an OSError/FileNotFoundError referencing the original, unexpanded name, since _write_git_file_and_module_config() looks for a directory using the pre-expansion name — but that error message itself does not contain the expanded/secret value.)
Control: the same call with an ordinary (no $) submodule name succeeds normally and creates .git/modules/<name> as expected — isolating the gap to submodule names containing environment-variable syntax specifically.
Already-known check
Checked all 35 published gitpython-developers/GitPython security advisories in full text. GHSA-hmq2-w58f-27jc covers the same name → .git/modules/<name> path but for path-traversal (../../..), not variable expansion. GHSA-59cr-6r3x-644w covers the sibling path field reaching unguarded os.makedirs(). GHSA-94p4-4cq8-9g67/GHSA-rwj8-pgh3-r573 cover expand_vars, but only on URL arguments. git log -p -S "separate_git_dir" -- git/repo/base.py shows neither fix commit touches this line; it is unchanged since its 2016 introduction (commit 57d05379).
The closest topical match is GHSA-8mcc-hrx5-hvxc ("clone_from()/clone() omit --separate-git-dir from unsafe_git_clone_options") — already fixed, and its own text discusses this same _clone() handling of separate_git_dir. That fix only added --separate-git-dir to the unsafe_git_clone_options denylist, which gates the direct-kwarg vector; Submodule._clone_repo() explicitly sets allow_unsafe_options=True for the gitfile-submodules path, unconditionally bypassing that exact denylist (confirmed: the reproduction above succeeds against current main, which already has that fix). The bug this report describes is independent of that denylist — the *value* itself is expanded via os.path.expandvars() regardless of the denylist outcome — so the two are distinct, but flagging the relationship explicitly since a reviewer would otherwise reasonably ask about it.
No advisory describes environment-variable expansion via the internal separate_git_dir path.
Suggested fix
Add expand_vars=False to the Git.polish_url(sep_dir) call in _clone(), matching the four other call sites already fixed for the same class of issue.
Verification level
Reproduced in a real environment: plain Linux, unmodified GitPython from current main, real git ≥ 1.7.5 behavior, an ordinary git submodule add + clone + submodule_update(init=True, recursive=True) workflow with no non-default configuration or monkeypatching required. Causation isolated via a control group (identical workflow, ordinary submodule name, no expansion).
AI Use Disclosure
AI (Claude) was used as a tool under my direction for code review, code summarization, and script development. A human reviewed and verified the AI-generated results before submission.
Credit: please use "baeseungwon1010" (github.com/baeseungwon1010).
Questions
- Does this need user interaction?
The CVSS vector sets UI:R. In practice the victim process must clone the malicious repository and run submodule_update with init=True. No special flag or config is needed.
- Is the secret sent to the attacker automatically?
No. The value is written as a directory name under .git/modules and, at DEBUG level, into a log line. The attacker has to be able to read the directory listing or the logs.
- Is GitPython 3.1.62 affected?
Yes. The affected range is 3.1.62 and earlier. Version 3.2.0 is the patched release.
References
How this page is made: the data panel comes from the GitHub Advisory Database, CISA’s KEV catalog, the NVD and FIRST EPSS, and updates twice a day. The analysis is written by an AI model from the sources listed above and checked by automated rules before it is published. Check versions against the vendor's advisory before you act. Methodology
More PyPI vulnerabilities
All PyPI entriesGHSA-23mf-xhv8-69c2
GitPython: arbitrary file read via attacker-controlled branch name in `Head.checkout()` (incomplete fix of CVE-2026-78679)
high7.1Oct 27, 2026GHSA-f9j4-qggq-h239
GitPython Remote.pull() refspec bypasses the --upload-pack unsafe-option guard, enabling arbitrary command execution
high8.1Oct 27, 2026GHSA-gq48-pqfc-9p58
GitPython: Submodule.move() follows an attacker-committed intermediate symlink outside the worktree
mediumOct 27, 2026GHSA-m64x-33q8-m5h7
GitPython parse_actor_and_date() is vulnerable to quadratic backtracking on malformed identity lines
high8.7Oct 27, 2026GHSA-w8jc-g24h-crhw
GitPython command wrappers omit unsafe protocol and option checks
high7.5Oct 27, 2026CVE-2026-87819
GitPython: Denial of Service via catastrophic backtracking (ReDoS) in Actor.name_email_regex — commit author/committer field parsing
high7.5Sep 30, 2026
