Skip to main content
Agent retrieves poisoned documents or vector store entries that redirect its objectives

​
How to resolve

Validate RAG data sources for integrity. Use multiple independent sources for cross-verification. Lock agent system prompts so goal priorities are explicit and auditable.

​
Risk

Governance/Compliance This vulnerability falls under ASI01:2026 — Agent Goal Hijack in the OWASP Top 10 for Agentic Applications. Attacker manipulates agent objectives or decision path via malicious input content. Security Agent retrieves poisoned documents or vector store entries that redirect its objectives. If exploited, this can compromise the agent’s integrity, confidentiality, or availability, potentially affecting downstream systems and data.

​
Explanation

Agent retrieves poisoned documents or vector store entries that redirect its objectives. This falls under ASI01 (Agent Goal Hijack): Attacker manipulates agent objectives or decision path via malicious input content.

​
Specifications

Trigger
  • Agent scan detects agent retrieves poisoned documents or vector store entries that redirect its objectives
The scanner looks for patterns such as:
  • RAG retrieval without content validation
  • vector store results fed directly to agent
  • unverified document sources in agent context
Severity
  • HIGH
Applies to: Agent