TrigGuard
TRIGGUARD CATEGORY

Execution authorization

Authorization before irreversible work—not after logs, tickets, or post-hoc audits. TrigGuard defines this category for automated systems.

Decisions bound to actions

Execution authorization is the process of evaluating whether an automated action should execute before it occurs. Unlike logging and monitoring, which happen after execution, authorization is a pre-execution control that determines whether the action is allowed at all. Every high-risk surface gets an explicit decision: PERMIT, DENY, or SILENCE, with evidence you can verify. See Decision model and Protocol spec.

Agents and CI don’t “mean well” by default

Automated systems can perform irreversible actions across CI/CD deployments, infrastructure changes, database migrations, payment execution, AI agent actions, and data exports. Speed without deterministic authorization is indistinguishable from incident risk. Authorization sits at the same choke point as the action—before commit. Architecture · Gate.

Execution request → decision

Execution request
→ POST /execute
→ Signal aggregation
→ TGSafetyEngine evaluation
→ Decision: PERMIT · DENY · SILENCE

Deterministic decision pipeline

The same request context under the same policy yields the same outcome, making behavior auditable and replay-safe. Callers describe intent; the engine evaluates deterministically; permitted paths are bound to verifiable receipts. Protocol · Runtime · API.

Prove what was authorized

Every decision produces a verifiable receipt that can be validated later for integrity, authenticity, and policy context. Receipts bind policy, request context, and outcome—Receipts, Verify, Verification docs, TG-RECEIPT-SCHEMA.

Gate, Verify, Control, SDK

Gate — execution authorization engine
Verify — cryptographic verification of receipts
Control — governance and policy enforcement
SDK — developer integration layer

Terraform deployment request

terraform apply
→ POST /execute
→ evaluation
→ PERMIT or DENY
→ execution receipt generated