SQL Injection in Snowflake CLI Input Validation

Improper input validation in Snowflake CLI versions prior to 3.27.0 allowed unsanitized user-controlled values to be interpolated into SQL strings that are executed as multi-statement queries. An attacker who is able to supply a malicious project configuration file or craft command-line input can cause Snowflake CLI to execute attacker-controlled SQL statements in the context of the victim's Snowflake session and active role. Successful exploitation requires either write or pull-request access to a project repository whose CI/CD pipeline runs Snowflake CLI under an elevated service account role, or the ability to supply untrusted input to CLI-wrapping automation. Impact is limited by the privileges held by the configured Snowflake role at execution time. The fix is available in Snowflake CLI version 3.27.0, which also addresses several additional security findings. Users must manually upgrade.

Casky was already ahead

This CVE exploits attack patterns that Casky's 323matched skills already investigate — long before this vulnerability was disclosed. Claude's reasoning model maps these techniques to MITRE ATT&CK, so practitioners who ran these skills have already seen the threat behaviour in their findings.

Analysis

CVE-2026-92903 represents a critical SQL injection vulnerability in Snowflake CLI versions before 3.27.0, where improper input validation allows attackers to inject malicious SQL commands through project configuration files or command-line arguments. This vulnerability is particularly dangerous because it executes arbitrary SQL statements within the context of the victim's Snowflake session and active role, potentially allowing unauthorized data access, modification, or deletion. Users running vulnerable CLI versions are at risk if they accept configuration files from untrusted sources or if their development environments are compromised. Organizations using Snowflake CLI for infrastructure-as-code, data pipeline automation, or CI/CD workflows face significant exposure, as attackers could leverage pull request access or configuration management systems to deliver malicious payloads.

Casky.ai's 323 mapped security skills leverage Claude's extended reasoning to identify attack patterns associated with TA0001 (Reconnaissance) and TA0043 (Resource Development) techniques that precede SQL injection exploitation. Practitioners using the platform would observe detection findings focused on: (1) suspicious input patterns in CLI argument parsing and configuration file loading that deviate from expected Snowflake query structures; (2) CWE-89 signature patterns indicating unvalidated string interpolation into SQL statements; (3) behavioral anomalies in SQL query composition that suggest dynamic statement construction from user-controlled sources; and (4) reconnaissance activity where attackers probe CLI behavior or enumerate configuration file formats before crafting injection payloads. The AI-driven analysis would surface these indicators before exploitation occurs, enabling teams to validate input sanitization, implement strict configuration file schemas, and enforce principle-of-least-privilege Snowflake roles for CLI operations.

Live Threat IntelligenceComing soon

Casky Risk Score

—

EPSS Probability

—

Shodan Exposure

—

GreyNoise Signal

—

Composite risk scoring from EPSS, CISA KEV, Shodan, and GreyNoise — 21 security APIs correlated into a single Casky Risk Score. Coming in Casky Pro. Join early access →

323 Casky skills cover this attack pattern

These skills use Claude AI's reasoning model to surface findings in the same attack categories as CVE-2026-92903.

Abuse of Public-Facing API: Mobile API Abuse

mobile security · low

Run

abusing-dpapi-for-credential-access

red teaming · high

Run

abusing-shadow-credentials-for-privesc

red teaming · high

Run

MITRE ATT&CK Techniques

TA0001 TA0043

Investigate the attack patterns behind CVE-2026-92903

Casky has 323 skills that investigate the attack patterns behind CVE-2026-92903. Run one and get CVSS-scored findings in 3 minutes.

Run the skill that detects this →

X Instagram LinkedIn