SQL Injection in Vector Database Collection Creation

PraisonAI before 4.6.78 fails to validate the caller-controlled dimension argument in the PGVector and Cassandra knowledge-store create_collection() backends. Although schema, keyspace, and collection-name identifiers are validated, the dimension value (declared as int but not enforced at runtime) is interpolated directly into the vector column of the generated CREATE TABLE DDL. A caller able to influence collection-creation dimensions can pass a string such as '3); DROP TABLE tenant_secrets; --' to inject SQL/CQL tokens into the statement executed by the database driver.

Casky was already ahead

This CVE exploits attack patterns that Casky's 322matched 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

PraisonAI versions before 4.6.78 contain a critical SQL injection vulnerability in the PGVector and Cassandra knowledge-store backends. The create_collection() function fails to properly validate the dimension parameter, allowing attackers to inject arbitrary SQL commands through what should be a numeric argument. While the application correctly validates schema, keyspace, and collection names, the dimension value is interpolated directly into CREATE TABLE DDL statements without enforcement of its declared integer type. This enables attackers to execute malicious SQL—such as dropping tables, exfiltrating data, or modifying database structure—by passing crafted strings like '3); DROP TABLE tenant_secrets; --' during collection creation. Organizations using PraisonAI for knowledge management, retrieval-augmented generation (RAG), or vector storage are at risk of data loss, unauthorized access, and system compromise.

Casky's 322 mapped skills leverage Claude's extended reasoning to detect the attack patterns associated with this vulnerability across MITRE ATT&CK techniques TA0001 (Reconnaissance) and TA0043 (Reconnaissance via reconnaissance tools and techniques). Practitioners using Casky will identify suspicious collection-creation requests that include SQL metacharacters in dimension parameters, anomalous DDL statements in database logs, and failed table creation attempts followed by unexpected data deletions. The platform's skills correlate input validation bypasses with database schema modifications, detecting the hallmarks of SQL injection attacks—particularly those targeting vector database operations. Findings will surface unvalidated user input flowing directly to SQL query construction, type-coercion failures at runtime, and the absence of parameterized query patterns in PraisonAI's vector backend code.

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 →

322 Casky skills cover this attack pattern

These skills use Claude AI's reasoning model to surface findings in the same attack categories as CVE-2026-60090.

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-2026-60090

Casky has 322 skills that investigate the attack patterns behind CVE-2026-60090. Run one and get CVSS-scored findings in 3 minutes.

Run the skill that detects this →

X Instagram LinkedIn