EASAPEnterprise agent assurance
Open referenceGitHub ↗Assess an agent
← All function guides Guide 01 · AI platform & engineering

Turn an agent build into an executable candidate.

Use EASAP to bind the exact system, declare its authority, and give every later test a stable subject.

BUILT FORAI platform engineers, application engineers, and quality teams
What you will leave with
✓Candidate manifest
✓Capability boundary
✓Scenario adapter
✓Regression gate
Your workflow
01

Describe the authority

Capture autonomy, exposure, data, scale, and actions. Treat this as the system contract, not a product description.

02

Freeze the candidate

Bind model, prompts, tools, memory, policy, connectors, and runtime to immutable digests so a result cannot drift underneath you.

03

Connect the boundary

Route tool requests through a mediated adapter. Undeclared effects must fail closed before they reach an external system.

04

Run and regress

Execute the generated scenario pack, retain traces and digests, and re-run the same gate whenever a material input changes.

Ready-to-use checklist
✓

Every capability has an owner and data scope

✓

The candidate can be reproduced from its manifest

✓

External effects are mediated and observable

✓

Failures have a deterministic replay path

HANDOFF TO THE NEXT FUNCTIONShare the assessment or probe ID, evidence requirements, open gaps, and scenario pack—not a screenshot.Revisit the EASAP model ↗