mr-onion

Phase 35 Security Research — Reviewer Questions (for Martin)

Companion to: phase-35-security-testing-research.md Date: 2026-10-13 Ask: Mark up / answer the questions below. You don’t need to read the full doc first — each question stands alone, with a pointer to the relevant section if you want context.

Background in one line: TestForge points at a website and runs reproducible test suites; Phase 35 adds a security area. “Standard” runs must be non-intrusive; “Comprehensive” runs may actively probe, behind a per-target authorisation gate. The questions below are the judgement calls where your experience matters more than mine.


A. The key judgement calls

  1. nuclei in Standard? Is nuclei’s default/safe template set genuinely non-intrusive enough to run in a Standard (non-intrusive) pass on an unauthenticated target — or should any nuclei run be Comprehensive-only? If some templates are Standard-safe, how would you draw that line (tags, severity, a curated allowlist)?

  2. ZAP passive/active boundary. We plan to run ZAP as a proxy behind the browser (passive) and optionally active-scan. Where exactly should passive end and active begin for our purposes? Anything ZAP calls “passive” that you’d still treat as intrusive?

  3. sqlmap — in or out of an automated product? Does automated SQLi testing belong inside a product at all, or should it be a manual escalation a human kicks off? If in, what guardrails are non-negotiable (scope limits, techniques to disable, risk/level caps)?

B. Standard vs Comprehensive placement

For each check, is our proposed placement right? (S = Standard/non-intrusive, C = Comprehensive/gated)

C. Open questions from the research

  1. ffuf vs gobuster as the single content-discovery primary — preference?
  2. ZAP integration shape — REST API + Automation Framework, the Playwright-proxy pattern, or both (proxy for passive, API for active)?
  3. Burp/Caido — agree these stay optional “bring your own proxy,” not a dependency?
  4. nmap overlap — nuclei ships network/SSL templates and we have testssl; is nmap still worth it for port/service discovery, or redundant for our web-app scope?
  5. Authorisation gate — what’s the minimum acceptable control before any active check runs against a target?

D. Anything missing?

  1. Any check, tool or whole category we’ve under-weighted for web-app DAST? Anything in the full doc’s “out of scope” list you’d argue back in?

Full detail, tool-by-tool evaluation and build order are in the companion research doc. Nothing here is locked — this is pre-decision input.