LIGERAAIAll skills ↗
Converts approved procedures into portable cases, graders, coverage and release thresholds.
Agent assurance · Evaluation · v3.2.0

SOP to Agent Evals

A portable evaluation suite that traces approved procedures to atomic requirements, cases, graders and release thresholds.

Reserve a pilot↗Read JSON record ↓
Problem this skill solves

Generic benchmarks do not show whether an agent follows the actual procedures, policies and failure controls that matter to a business.

Best moment to use it: Use it when a real procedure, policy or known failure must become a portable agent test suite.

How it works

What the agent actually does

  1. 01

    Trace

    Turns approved SOP clauses, policies and known failures into atomic source-linked requirements.

  2. 02

    Design

    Creates normal, edge, refusal and recovery cases with the appropriate graders.

  3. 03

    Set gates

    Measures coverage gaps and defines thresholds the release process can apply.

What you receive

Concrete, reviewable outputs

  • Source and requirement register
  • Capability and safety test cases
  • Deterministic, trace, rubric or human graders
  • Coverage report and release thresholds
Example in practice

A refund SOP becomes 28 traceable test cases, including missing evidence, limit breaches and mandatory human escalation.

What it does not do

It defines and validates the evaluation package; it does not claim the target agent passed until an authorized harness runs it.

Buyer

AI product, QA, operations and risk teams

Business result

A portable evaluation suite that traces approved procedures to atomic requirements, cases, graders and release thresholds.

Release stage

Validation release

Stripe Checkout remains disabled until payment, webhook and fulfillment configuration are verified. Final scope and delivery are confirmed before activation.

Purchase path

Pilot request

Pricing

$249 USD

Individual license · or $99/month with updates.

Compatibility

Agent Skills-compatible workspaces · ChatGPT · Codex · OpenAI API workflows

Trust record

Version 3.2.0

Last catalog verification: 2026-08-30. Evidence-first workflow with explicit human approval boundaries.

Commercial terms

Clear license.
Human accountability.

Individual commercial-use license; redistribution and resale require a separate agreement.

Ask about team or enterprise licensing ↗