Public references

Rules mapped to public security guidance.

Our guardrails line up with the public guidance written for AI agents. A reviewer can read the overlap in one place, with the frameworks we map against named plainly.

Sill holds no security certification today. This page is a reference, not a certificate. You remain responsible for your own compliance.

Mappings

Control by control, line by line.

GUARDRAILS YOU CONFIGURE

Sill guardrails mapped to OWASP LLM Top 10, OWASP Agentic Top 10, MITRE ATLAS, and NIST AI RMF
CONTROLOWASP AGENTIC 2026OWASP LLM 2026MITRE ATLAS 2026.10NIST AI RMF
Allowlisted agents onlyASI03, ASI10—Initial AccessManage
Require a signed order from the customerASI01, ASI03LLM06—Manage
60 calls per minute, per agentASI08LLM10ImpactManage
100 calls per minute, per IPASI08LLM10ImpactManage
Per-transaction spend capASI01LLM06, LLM10ImpactManage
Daily spend cap, per customerASI08LLM06, LLM10ImpactManage
Human review for refunds and cancellationsASI08, ASI09LLM06ImpactManage
No fake-urgency ordersASI09 (partial)———
No drip pricingASI09 (partial)———
Block instruction-override textASI01LLM01AML.T0051 (partial)Measure
Country allow / deny——Initial AccessManage
Site-wide rate capASI08LLM10ImpactManage
Cart stays under the customer’s budgetASI01LLM06—Manage
Cart currency matches the authorizationASI01LLM06—Manage
One customer, one customer’s dataASI03LLM06ExfiltrationManage
Approved agent toolsets onlyASI02, ASI04 (partial)LLM03 (partial)AML.T0011.002 (partial)Measure
No silent subscriptionsASI09LLM06—Manage
Block invisible-character attacksASI01LLM01AML.T0051 (partial)Measure
Block orders that contain secretsASI02 (partial)LLM02 (partial)AML.T0098 (partial)Measure
Short-lived order authorizationsASI03LLM06EvasionManage
Order payload size limit—LLM10——
Emergency kill switchASI08LLM10ImpactManage
Budget cap, per customer organizationASI08LLM06, LLM10ImpactManage
Your own rule (advanced)merchant-definedmerchant-definedmerchant-defined—

ALWAYS ON

Sill always-on protections mapped to OWASP LLM Top 10, OWASP Agentic Top 10, MITRE ATLAS, and NIST AI RMF
CONTROLOWASP AGENTIC 2026OWASP LLM 2026MITRE ATLAS 2026.10NIST AI RMF
Signature check on every signed requestASI03, ASI10—Initial Access, Evasion; AML.T0096Manage
Block replayed ordersASI03—EvasionMeasure
Lock out after repeated auth failures—LLM10Credential AccessManage
Reject a request bound to another siteASI03—EvasionManage
Anti-fingerprinting on identity reads——Discovery—
Signed webhooks only———Manage
Evaluation budgets with a safe default when exceededASI08LLM10ImpactManage
Tamper-evident, linked audit chainASI06 (partial)——Measure, Manage
Personal data redacted at the storage layer—LLM02 (partial)ExfiltrationManage
Scrub agent-visible responses—LLM02, LLM05, LLM07——
Delegation chain verified before trust is grantedASI07 (registered-agent)——Manage
Signed cart cannot be editedASI01LLM06—Manage

OUT OF SCOPE

Risks that come from training or hosting a model
Sill does not train, host, or retrieve for models.
Unexpected code execution
Sill is an authorization layer, not an execution sandbox.
Agent-to-agent delegation outside a registered agent
Delegation from a registered agent is covered. Chains that start outside a registered agent are outside what Sill verifies.
The agent’s own runtime, host or build chain
Those live inside the agent, outside the request Sill sees.
Check your site

Is your business open for agents?

Type your domain. See your site the way an AI assistant reads it.

Is your business open for agents?

Type your domain. See your site the way an AI assistant reads it.

Check your site

Is your business open for agents?

Type your domain. See your site the way an AI assistant reads it.