Security Questionnaire
A questionnaire is an evidence pack, not a marketing page. Answer what is true today. Mark gaps. Do not invent certifications, subprocessors, or controls.
For a code/design vuln hunt, use security-review. For a STRIDE model, use threat-modeling. For privacy DSAR process, use privacy-request-playbook.
Workflow
- Identify the pack (SIG, CAIQ, custom RFP, SOC 2 bridge) and the product/environment in scope.
- Collect facts the user already has: hosting, IdP, encryption, logs, backups, subprocessors, certs (type + date), pentest window.
- Answer each question in one of: yes + evidence, no + plan, partial + boundary, N/A + why.
- Point to artifacts: policy name, ticket, SOC report section, config screenshot description — never attach secrets.
- Flag contradictions (we say SSO required, signup is password-only).
- Produce a gap list the sales/security owner can actually close.
Output format
## Questionnaire: <pack / customer>
**Scope / env:** …
**Evidence used:** …
### Answers
| ID | Question gist | Answer | Evidence | Residual |
### Gaps to close before send
1. …
### Do not claim
…
Rules
- Never invent SOC 2 / ISO / HIPAA / FedRAMP. If the letter is not in hand, say "not certified" and describe the control you do run.
- Do not paste real secrets, customer data, or internal IPs.
- "Yes" without a pointer is a draft, not a sendable answer.
- Scope tightly: production SaaS ≠ employee laptops ≠ a mobile app unless asked.
- Subprocessors: only list what the user named. Missing list is a gap, not a guess.
- Future tense ("we will encrypt") is a roadmap, labeled as such.
- If the question is legal (DPA, liability), draft a factual control answer and mark it for counsel — do not give legal advice.
Edge cases
- Blank company: produce the question list grouped by theme and the evidence to gather; do not fabricate a security program.
- Conflicting answers in an old pack: surface the conflict; do not silently pick the nicer one.
- Pentest / vuln details: summarize residual risk; do not paste exploit steps into a customer-facing cell.
- User wants a threat model: point at
threat-modelingand keep this file to Q&A.
---