Skip to content

SanityOps Framework ​

The root of the SanityOps methodology — the governance foundation that defines what SanityOps is, how the closed loop works, and how the three downstream subsets (Inspect, Risk, Quality) connect.


Why This Section? ​

SanityOps is an open, vendor-neutral methodology for AI Agent governance, built on logic-artifact defect inspection and extending into Risk scanning and Quality assessment. This section holds the normative core of the framework itself — before you dive into the three professional subsystems and the surrounding ecosystem.


Framework Documents ​

  • Framework Overview ​

    The complete whitepaper — what SanityOps is, why it exists, and how everything fits together.

    A 14-section end-to-end narrative covering the three enterprise challenges, the root cause, the three professional systems, lifecycle integration, boundaries, outputs, and FAQ. Start here for the big picture.

  • Core Specification ​

    The normative foundation of the framework.

    Defines the governance closed loop, the unified object and artifact model, terminology and numbering rules, the Ratchet Mechanism, and release gates.

  • Relevance ​

    The impact-mapping specification.

    Maps defects discovered during inspection to their downstream impact on Risk and Quality, completing the governance feedback loop.

  • Try the Tools ​

    Openness overview and tooling entry points.

    How to access the open-source lite inspection skill and Inspect CLI (Inspect + Risk Explicit), the Risk Implicit and Quality SaaS demos, and self-hosted deployment options.

  • FAQ ​

    Frequently asked questions.

    Answers to common questions for AI service managers, developers, operators, and security teams, with sources cited from the official specifications.

  • Samples ​

    Worked examples from the tools.

    Illustrative Logic Artifacts and defect inspection reports produced by the SanityOps tools, showing findings, risk validation rounds, and remediation tracking.


How the Framework Connects ​

Framework (Core + Overview + Relevance)   ← You are here
    │
    ├─ Inspect   (defect discovery)
    ├─ Risk      (explicit + implicit security)
    └─ Quality   (service quality & reliability)

The Framework is the spine. Core defines the rules, Relevance links findings back to impact, and Overview gives you the map. From here, proceed to Inspect → Risk → Quality, then explore the ecosystem in Compare.


A recommended reading path through the core documents. You can also jump directly to any section that interests you.

1. Overview ​

Start here for the big picture — what SanityOps is, why it exists, and how the professional systems fit together.

→ Overview

2. Core Framework ​

Understand the governance closed loop, the Ratchet Mechanism, release gates, and the framework's structural foundation.

→ Core Specification

3. Inspect ​

Learn how to statically scan System Prompts, Skills, and Tool Schemas for definition defects and boundary issues, then check whether granted permissions are proportionate to assigned responsibilities.

Note: Run Inspect Permission last, after the other Inspect sub-specifications, and subject to the Gate-0 precondition.

→ Inspect Introduction

4. Risk ​

Audit explicit risks in logic artifacts, then validate implicit vulnerabilities through dynamic attack simulation in shadow sandboxes.

→ Risk Introduction

5. Quality ​

Quantitatively assess user-visible output quality and task reliability — the final gate before production release.

→ Quality Introduction

6. Relevance ​

Map defects discovered during inspection to their potential impact on risk and quality, completing the governance feedback loop.

→ Relevance

7. Compare ​

Understand how SanityOps relates to adjacent tools and evaluation frameworks in the broader ecosystem.

→ Compare

If you are a...Start with
First-time readerOverview
Agent developerInspect → Risk
Security auditorRisk Explicit → Risk Implicit
QA engineerQuality RAG-Agent or Quality Tool-Agent
Platform / CI/CD engineerCore Specification