AGENTFORCE-OG-01: OGP Metadata Prompt Injection
🔴 High · OpenGraph Security
Detects prompt injection patterns embedded in Open Graph Protocol metadata (og:description, og:title) on URLs referenced by agent actions, MCP tools, or prompt template data providers. Headless agents process these tags without human visual inspection, making OGP a stealth injection channel.
Details
| Field | Value |
|---|---|
| Rule ID | AGENTFORCE-OG-01 |
| Severity | High |
| Category | OpenGraph Security |
Remediation
Review and sanitize the og:description on the MCP server endpoint before connecting agents. Apply input validation to all OGP metadata before it enters the LLM context window.