Field manual · AI for authorized security testing · Edition November 2026 · Updated 2026-11-18

Phase · November 2026

AI for pentest scoping: the best model, and how to use it

The best AI for pentest scoping is Kimi K3.1 Pro, because a statement of work becomes a test plan by reasoning, and it is the strongest reasoner here.

Every API model holds 1,048,576 tokens, so judge the plan by whether it stays inside the assets and rules you were given.

Updated November 18, 2026Authorized work only
01Method

What a scope has to settle before anyone tests

Pre-engagement is the first stage in the PTES, and planning is the first phase in NIST SP 800-115. Name what is in (hosts, applications, accounts, dates), what the statement of work never names, the hours and rate, and the evidence each later phase brings back. If there is no exclusion list, say so.

Rules of engagement

Only on systems you own, a signed scope, a CTF or a lab. A model that answers is not permission.

02With a model

Hand the model the statement of work

Paste the statement of work and ask for a plan. Set reasoning effort to max. (reasoning) Kimi K3.1 Pro is $3.00 in and $8.00 out per 1M tokens. PentAGI can store the plan.

Prompt

Here is the statement of work for an authorized test. List in-scope assets, exclusions, dates and rules, then the evidence each phase should produce. Do not add targets that are not in the statement of work. Do not write exploit code.

Next is recon or web applications. The model page is Kimi K3.1 Pro. Other grades are on the home matrix.