OWASP Agentic Top 10 Coverage

How Strathon maps to the OWASP Top 10 for Agentic Applications 2026, the threat model Strathon's design is anchored on.

The OWASP Top 10 for Agentic Applications (ASI01–ASI10) is the reference taxonomy for how agentic systems get attacked: goal hijacking, tool misuse, privilege abuse, supply-chain compromise, and the rest. Strathon ships policy templates mapped to these threats, available via GET /v1/policy-templates and applicable with a single API call.

A note on honesty up front: not every item on this list is fully solvable at the tool-call boundary, and the table below says so where it matters. Some threats (memory and context poisoning, cascading failures) are addressed through detection and containment rather than a single pre-execution verdict, and a few have legs that live outside any runtime firewall. The Scope & Limitations page is the long version of where those lines fall.

Threat coverage

OWASP ThreatTemplateStrathon mechanism
ASI01 Agent Goal Hijackprompt-injection-detectionCEL policy on span attributes
ASI02 Tool Misuse and Exploitationtool-access-allowlistDeny-by-default (allow-list mode)
ASI03 Identity and Privilege Abuse(built-in)Scoped API keys, RBAC, MFA, per-key rate limits
ASI04 Agentic Supply Chain Vulnerabilities(built-in)MCP gateway with policy evaluation, egress proxy, credential scanning
ASI05 Unexpected Code Executiontool-access-allowlistBlock/allow-list on shell, code, and SQL tools; approval before code execution
ASI06 Memory and Context Poisoning(built-in)Behavioral drift detection (Vigil), halt propagation, content redaction
ASI07 Insecure Inter-Agent Communication(built-in)MCP gateway policy evaluation, fail-closed enforcement
ASI08 Cascading Failuresiteration-budget-guard, cost-budget-guardBudgets with auto-halt, circuit breakers, kill switches, halt propagation
ASI09 Human-Agent Trust Exploitation(built-in)Human approval workflows, tamper-evident audit log, SARIF export
ASI10 Rogue Agents(built-in)Vigil drift detection, heartbeat monitoring, kill switches

Where the boundary is strongest

ASI02, ASI05, and ASI07 are the threats a tool-call firewall addresses most directly: each is a question about whether a specific tool call should run, and that is exactly the decision Strathon makes before the call executes. Deny-by-default allow-lists, argument-level CEL rules, and approval gates give you a hard stop at the moment an agent's decision becomes a real action.

ASI03 and ASI09 are addressed through identity and accountability primitives (scoped keys, RBAC, multi-party approval, a tamper-evident audit trail) rather than per-call matching alone.

Where it is detection and containment, not prevention

ASI06 (memory and context poisoning) and ASI08 (cascading failures) produce calls that carry no in-band signal at the moment they fire. Strathon addresses these through behavioral drift detection, budgets with auto-halt, circuit breakers, and halt propagation, which surface and contain the effect rather than blocking a single malicious call. This is real coverage, but it is a different mechanism than pre-execution enforcement, and worth understanding as such.

Applying the templates

Each template above maps to a ready-made policy. List them with GET /v1/policy-templates, then apply one to a project with a single call, or create the policy in the dashboard from the templates picker. Templates are a starting point: tune the CEL match and action to your own tools and risk tolerance before relying on them.

Related