Penetration testing rules of engagement: a template

#Template

# RULES OF ENGAGEMENT — [Organization] / [Tester]
Version: 1.0           Effective: [YYYY-MM-DD] to [YYYY-MM-DD]

1. AUTHORIZATION
   Authorizing officer: [name, title]   Asset owner sign-off: [list]
   Third-party consents (cloud, SaaS, MSSP): [attached / not required]

2. OBJECTIVES
   - [e.g. reach a canary record in billing DB from the internet]
   - [e.g. validate segmentation between corp and CDE]

3. SCOPE (machine-readable copy attached as scope.yaml)
   In scope:  [domains, CIDRs, cloud accounts, apps + roles, LLM endpoints]
   Excluded:  [hosts, paths, third parties, production DBs for write]
   Accounts:  [test users per role, MFA arrangements]

4. WINDOWS & RATE LIMITS
   Testing windows: [24/7 | business hours TZ]   Max req/s per host: [n]

5. TECHNIQUES
   Allowed:   [exploitation, credential attacks on test accounts, phishing: Y/N]
   Forbidden: [DoS, destructive payloads, data modification, physical: Y/N]
   Human approval required for: [list]

6. DATA HANDLING
   Proof of impact via canary records: [locations]
   Incidental data: redact before storage; retention [n days]; encryption [..]

7. STOP CONDITIONS & KILL SWITCH
   Stop immediately if: [service degradation, out-of-scope access, real
   attacker activity discovered]   Kill switch owner: [name, phone]

8. COMMUNICATIONS
   Daily status: [channel]   Critical finding: notify within [n] hours
   SOC informed: [yes / no — blind test]  Deconfliction contact: [name]

9. REPORTING
   Findings format, severity model, retest policy, report recipients

10. SIGNATURES

#Fields autonomous testing adds

• Machine-readable scope enforced at the network layer, versioned with the RoE.
• Canary plan so impact is proven without touching real data.
• Action-level approval rules — which techniques need a human tap.
• Audit log destination so your SOC can watch the test in real time.