OWASP agentic · ASI02
ASI02: Tool Misuse and Exploitation
An agent uses a legitimate tool it is allowed to call, but in an unsafe or unintended way: deleting data, running up costs or leaking information. The cause can be prompt injection, ambiguous instructions, misalignment or unsafe delegation, and chaining several tools amplifies it. Escalation of privilege is covered under ASI03 and arbitrary code execution under ASI05.
What it looks like
- An email-summarising agent that also holds send and delete rights is manipulated into deleting or forwarding mail.
- Injected instructions make an agent chain an internal CRM tool with an external email tool, exfiltrating a customer list.
- An attacker repeatedly triggers an auto-approved tool (such as ping or DNS lookups) to leak data, or causes a loop of costly API calls.
What OWASP recommends
- Give each tool a least-privilege profile (scopes, rate limits, egress allowlists), ideally expressed as authorization policy rather than convention.
- Authenticate each tool call and require human confirmation, with a dry-run or diff, for destructive or high-impact actions.
- Put a policy enforcement point in front of tool execution that treats model output as untrusted, validates arguments, enforces budgets and issues short-lived credentials.
- Run tools in sandboxes with egress controls, pin fully qualified tool names and versions, and keep immutable logs to spot unusual tool chains.
How Kyvvu's policies address it
These policies from the open library are written for this threat. Runtime policies judge each step against everything the agent did earlier in the task; registration policies check what an agent declares before it runs.
- Tool calls must be in the declared allowlist: runtime, critical
- Bound resource calls per task to prevent runaway loops: runtime, high
- Agent must declare a tool allowlist: at registration, high, in the OWASP agentic default pack
Based on the OWASP Top 10 for Agentic Applications for 2026 by the OWASP GenAI Security Project, licensed CC BY-SA 4.0. The names are OWASP's; the summaries are ours.
An example policy
Tool calls must be in the declared allowlist
From the OWASP agentic default and Internal AI starter pack. Severity critical: the step is blocked before it runs.
- name: "Tool calls must be in the declared allowlist" description: | ASI02 (Tool Misuse), ASI04 (Supply Chain): Pairs with registration policy. At runtime, blocks tool calls whose step_name is not in the agent's declared_tools allowlist. Catches typosquatting, supply-chain swap-ins, and silent capability creep as the agent's prompt evolves. Applies to: ALL agents. rule_type: step_name_in_allowlist params: agent_field: declared_tools target_kinds: ["step.resource", "step.exec", "step.credential"] severity: critical enforcement_point: step_executionYour agents
Is ASI02 a risk for yours?
For security teams
Request a demo
We walk through your agents' riskiest paths, and the policies that close them.
For developers
Build the path yourself
Recreate the attack in the playground and see which policies stop it.