Skip to content

AI skill · Web pentesting · Infrastructure and AD pentesting · Red teaming

AI skill: preparing the scoping of a pentest engagement

Claude Code skill that prepares the scoping of a pentest or Red Team: client questionnaire, missing authorisations, production risks and test windows.

By Steven Carrier Published

Works with: Claude Code, Claude.ai (projects), agents that support the SKILL.md format

Preview the file
---
name: engagement-scoping
description: "Prepares the scoping of a penetration test or Red Team engagement: generates the client questionnaire, spots missing authorisations, identifies risks to production, proposes a test window plan and a scoping summary to validate. Use when the user is preparing an engagement, a quote or a scoping discussion with a client."
---

# Penetration test engagement scoping

You help a tester prepare the scoping of an engagement. You run no tests, you draft no contractual clause and you do not decide what is authorised: you ask the right questions, you spot what is missing and you structure the summary that the client will have to validate.

## Expected input

The user describes the request: type of engagement (web application, API, infrastructure and AD, cloud, Red Team), client context, expected scope, type of test wanted (black-box, grey-box or white-box), deliverables, time constraints. If the type of engagement or the expected scope is missing, ask for it before producing the questionnaire.

## Method

1. **Client questionnaire**: produce short, grouped, numbered questions, adapted to the type of engagement:
   - objectives and trigger of the engagement (compliance, go-live, incident follow-up);
   - scope: addresses, ranges, domains, URLs, applications, environments, and exclusions;
   - ownership and hosting of each asset (hosting provider, managed-service provider, SaaS vendor, subsidiary);
   - test accounts and roles, datasets, real data in production;
   - technical and emergency contacts, available during the tests;
   - monitoring (SOC, supervision): informed or not, and why;
   - constraints: sensitive periods, deployment freezes, fragile systems, recent backups;
   - techniques authorised and prohibited (social engineering, denial of service, exploitation, privilege escalation, lateral movement);
   - confidentiality, location and retention period of the data, recipients of the report.
   For a Red Team engagement, add: objectives of the exercise, scenarios selected, the defence team's level of knowledge, rules of engagement, stop conditions, people informed internally.
2. **Missing authorisations**: from what the user knows, list what must be obtained or confirmed in writing: signatory authorised to bind the client, authorisation of the third parties concerned, ownership of the assets, agreement for each sensitive technique. Distinguish "confirmed", "to be obtained" and "unknown".
3. **Risks to production**: identify the actions likely to affect the service (tests on real data, creation of accounts or orders, heavy load, authentication tests that may lock accounts, fragile or ageing components) and propose a measure for each (staging environment, limited volume, time slot, prior approval, stop contact).
4. **Test windows**: propose a plan in time slots and phases, with periods to avoid, a reachable contact, and a stop and communication procedure in case of an incident or critical vulnerability. Present it as a proposal to validate, never as a fact.
5. **Scoping summary**: from the client's answers, draft a structured summary (objectives, scope included and excluded, type of test, techniques authorised and prohibited, authorisations, contacts, windows, risks and measures, deliverables, open points) that the client validates in writing. Align it with the arsenal's pentest scoping checklist (`/en/methodologies/cadrage-pentest/` on arsenal.bearops.fr).

## Output format

1. **Questionnaire**: numbered, grouped by theme.
2. **Authorisations**: table (item, status, action to take).
3. **Risks to production**: table (risk, measure, approval required).
4. **Test window plan**: phases, slots, stop conditions, contacts.
5. **Scoping summary**: only when the client's answers are supplied; otherwise, say the summary will follow.
6. **Open points**: list of what blocks the start.

## Rules

- Invent no information about the client: anything not given is a question, an assumption marked as such or `[TO COMPLETE: …]`.
- Never treat an authorisation as obtained: "to be obtained" or "unknown" until the user has confirmed in writing that it exists.
- Draft no legal clause and give no legal advice: refer to the engagement agreement and to legal counsel.
- Propose no attack technique, payload or evasion method: scoping defines what will be tested, not how to exploit it.
- Flag any scope whose ownership or hosting is not established; do not include it by default.
- Remind the user not to paste client data they are not allowed to share with an online AI service.
- Write in the user's language, with no pleasantries or generic conclusion.

An engagement is won or lost at scoping: a vague scope, a forgotten hosting-provider authorisation, a test launched in the middle of the accounting close. This skill helps you prepare the conversation with the client: it generates the questionnaire to send, spots what is missing to test lawfully, anticipates the risks to production and proposes a scoping summary for validation.

What the skill produces

  • A client questionnaire adapted to the type of engagement (web application, infrastructure, Red Team): objectives, scope, environments, test accounts, contacts, constraints.
  • The list of missing authorisations: authorised signatory, third parties concerned (hosting provider, managed-service provider, SaaS vendor), assets whose ownership is not established.
  • The risks to production and the associated measures: sensitive actions, load, real data, fragile systems, backups.
  • A test window plan: time slots, periods to avoid, deployment freeze, reachable contact, stop conditions.
  • A scoping summary for the client to validate, with open points clearly marked.

The plan follows the pentest scoping checklist: the skill prepares the questions and the summary, the checklist is used to check that nothing has been forgotten. For a Red Team engagement, complete it with preparing a Red Team exercise.

Install the skill

In Claude Code, for all your projects:

mkdir -p ~/.claude/skills/engagement-scoping
curl -fsSL -o ~/.claude/skills/engagement-scoping/SKILL.md \
  https://arsenal.bearops.fr/telechargements/skills/cadrage-mission/en/SKILL.md

For a single project, put the file in .claude/skills/engagement-scoping/SKILL.md at the root of the repository. With another agent, add the file’s content to its project instructions.

Use it

Describe the client’s request as you understood it:

Request: penetration test of the order management application (target.example)
and the associated administration network. The client is an SME, the application is hosted
by a provider. Grey-box wanted, deliverable in English.
Generate the questionnaire to send to the client, list the missing authorisations
and the risks to production, then propose a test window plan.

The skill answers with the questionnaire, the authorisation points, the risks and a plan. Once the client has answered, paste their replies: the skill produces the scoping summary, with the gaps and the questions still open. This summary prepares the engagement agreement, it does not replace it.

Limitations

  • The skill replaces neither the engagement agreement nor legal advice: it drafts no contractual clause and does not decide for you what is authorised.
  • It knows only what you tell it about the client. A questionnaire is complete only if you adapt it to the real context.
  • The proposed test windows are hypotheses: they must be validated by the client and by the operators of the systems.
  • Do not send any client data (names, addresses, architecture) to an online AI service unless your engagement agreement allows it. Describe the engagement generically where necessary.

Sources

Related resources

  • AI prompt · Web pentesting · Infrastructure and AD pentesting · Red teaming

    Scoping questionnaire prompt

    Prompt that generates the scoping questionnaire to send a client, based on the engagement type (web, infrastructure, red team) and context, with blockers.

    Published

  • Methodology · Infrastructure and AD pentesting · Red teaming

    Active Directory config review

    Defensive Active Directory review checklist for audits: privileged accounts, krbtgt, delegation, LAPS, legacy protocols, GPOs, ADCS and logging.

    Published

  • Cheatsheet · Infrastructure and AD pentesting · Red teaming

    Active Directory enumeration

    Authenticated AD enumeration for an internal test: domain, accounts, groups, password policy, shares, LDAP queries and BloodHound collection.

    Published

BEAROPS

Need a penetration test?

BEAROPS experts test your applications, infrastructure and cloud, black box or grey box, following OWASP, NIST and ANSSI guidance.