# The instruments

> GAMMA-Ray v0.0.2 (draft). The instruments GAMMA-Ray reads: what each is, who published it, what it rates, its licence, when it was read and its caveats.
> Page: https://gamma-ray.sgit.ai/instruments/ · Source: content/instruments/index.md


Standards, frameworks, a protocol specification and a law. They are different kinds of thing: a law binds, a framework is adopted by choice, a vendor's article is neither. Each is kept whole and cited; none is merged into another.

### Salesforce Agentic Maturity Model

| | |
|---|---|
| Publisher | Salesforce |
| Kind | vendor article |
| Date | 2025-04-10 |
| Rates | the agent's use case |
| Source | [www.salesforce.com/news/stories/agentic-maturity-model/](https://www.salesforce.com/news/stories/agentic-maturity-model/), read 2026-11-09 |
| Licence | Salesforce's; quoted briefly with the source |
| Provisions here | 34, of which a question reaches 31 |

**Caveat.** A vendor's news story with no scoring rubric. Its lists of how to advance are advice, not tests.

### Enterprise Agentic AI Maturity Model (EAAMM)

| | |
|---|---|
| Publisher | S. Emelyanov, ICAIMT 2026 |
| Kind | conference paper |
| Date | 2026-04-22 |
| Rates | the agent's use case |
| Source | [www.icaimt.com/2026/agentic-ai.html](https://www.icaimt.com/2026/agentic-ai.html), read 2026-11-09 |
| Licence | the author's; quoted briefly with the source |
| Provisions here | 12, of which a question reaches 12 |

**Caveat.** One author; a convenience sample of 136 who rated themselves. Its five levels are the five user roles of Feng, McDonald and Zhang, Levels of Autonomy for AI Agents (arXiv 2506.12469), which we have not yet read at source.

### State of Agentic AI Security and Governance 2.01

| | |
|---|---|
| Publisher | OWASP GenAI Security Project |
| Kind | open report |
| Date | 2026-06-01 |
| Rates | the deployed agent (tier) and the organisation (governance) |
| Source | [genai.owasp.org/resource/state-of-agentic-ai-security-and-governance/](https://genai.owasp.org/resource/state-of-agentic-ai-security-and-governance/), read 2026-11-10 |
| Licence | CC BY-SA 4.0 |
| Provisions here | 35, of which a question reaches 34 |

**Caveat.** Its matrix is described by OWASP as a self-assessment and roadmapping tool, not a binding compliance verdict.

### Model Context Protocol, Authorization (draft)

| | |
|---|---|
| Publisher | Model Context Protocol project |
| Kind | protocol specification |
| Date | draft, read 2026-11-09 |
| Rates | the agent's published endpoint |
| Source | [modelcontextprotocol.io/specification/draft/basic/authorization](https://modelcontextprotocol.io/specification/draft/basic/authorization), read 2026-11-09 |
| Licence | the project's; requirements quoted with the source |
| Provisions here | 15, of which a question reaches 15 |

**Caveat.** A draft page. Fifteen requirements chosen by us as a first cut, not the whole specification. It applies only where an endpoint of this kind is published.

### Agentic AI Governance Maturity Model (AAGMM)

| | |
|---|---|
| Publisher | V. Acharya, arXiv 2604.16338 |
| Kind | preprint |
| Date | 2026-03-13 |
| Rates | the organisation |
| Source | [arxiv.org/abs/2604.16338](https://arxiv.org/abs/2604.16338), read 2026-11-09 |
| Licence | CC BY-NC-ND 4.0: cited by domain name only, not adapted |
| Provisions here | 12, of which a question reaches 12 |

**Caveat.** One independent author; validated by simulation (750 runs), not in production. Only its twelve domain names are used here.

### Agentic AI adoption maturity model

| | |
|---|---|
| Publisher | Microsoft Learn |
| Kind | vendor guidance |
| Date | read 2026-11-09 |
| Rates | the organisation |
| Source | [learn.microsoft.com/en-us/agents/adoption-maturity-model/](https://learn.microsoft.com/en-us/agents/adoption-maturity-model/), read 2026-11-09 |
| Licence | Microsoft's; quoted briefly with the source |
| Provisions here | 88, of which a question reaches 45 |

**Caveat.** A self-assessment that refers to Microsoft products. The 88 items are the bullets of its quick reference table, as the page served them.

### AI Risk Management Framework 1.0, subcategories

| | |
|---|---|
| Publisher | NIST |
| Kind | framework |
| Date | 2023-01-26 |
| Rates | the organisation and the AI system |
| Source | [airc.nist.gov/airmf-resources/playbook/](https://airc.nist.gov/airmf-resources/playbook/), read 2026-11-10 |
| Licence | a United States government work; freely reusable |
| Provisions here | 72, of which a question reaches 50 |

**Caveat.** About AI in general, not agents. The 72 subcategories are taken from NIST's playbook data file.

### Regulation (EU) 2024/1689 (the AI Act), selected articles

| | |
|---|---|
| Publisher | European Union |
| Kind | law |
| Date | 2024-06-13 |
| Rates | the AI system, its provider and its deployer |
| Source | [eur-lex.europa.eu/eli/reg/2024/1689/oj](https://eur-lex.europa.eu/eli/reg/2024/1689/oj), read 2026-11-10 |
| Licence | EU law; freely reusable |
| Provisions here | 11, of which a question reaches 10 |

**Caveat.** Eleven articles chosen by us, by title only. Most bind only for high-risk systems. A later amending regulation exists and is not reflected. This is not a legal reading.

## Not an instrument: the lens

RiskMandate's Agent Behaviour Policy (mandate, grant, delta, barrier) is how GAMMA-Ray measures, not something it maps. See [the lens](../lens/index.md).

## Read, and not yet in the graph

- **AgentStripes** (arXiv 2610.18603): an evidence-graded maturity model for one released agent, from its repository; 40 capabilities in 12 areas. Its rule that the evidence level "is not chosen by the judges" is the model for ours.
- **SEI and Accenture, AI Adoption Maturity Model** (8 June 2026): eight dimensions, five levels, about AI adoption in general.
- **Cloud Security Alliance, Agentic AI Governance Maturity Model**: a draft of 27 March 2026, with a 22-question self-assessment.
- **Feng, McDonald and Zhang, Levels of Autonomy for AI Agents** (arXiv 2506.12469): the source of the five user roles.