← ClaudeAtlas

securitymaxxlisted

Score a shipped app against a 10-point security checklist with cited file:line or live-probe evidence, then emit a designed HTML scorecard. Unverifiable items score zero, so there are no hallucinated passes. Use when asked to audit an app's security, check whether something is safe to launch, review a vibe-coded or AI-generated codebase before real users touch it, or produce a security scorecard. Trigger — /securitymaxx <repo-or-url>, "audit this app's security", "is this secure enough to ship", "security scorecard".
whereisdotva-gif/securitymaxx · ★ 0 · Code & Development · score 70
Install: claude install-skill whereisdotva-gif/securitymaxx
# securitymaxx Score one surface 0–10 against the checklist below. Every verdict carries evidence you actually read or ran **this session**. Produce an HTML scorecard. The whole point is the evidence rule. An audit that says "looks fine" is worth nothing, and an LLM will happily produce ten of those in four seconds. Here, a PASS you cannot cite is not a PASS — it is UNKNOWN, and UNKNOWN scores zero. That single rule is what makes the number mean something. ## Authorization — read this before running Run this against **systems you own, or systems you have written authorization to test.** The skill is read-only by design: - Static analysis: read and grep source, config, and built output. - Live probes: ordinary HTTP requests only — headers, redirect chains, fetching a public bundle the browser would fetch anyway. - **Never**: exploit attempts, credential testing, brute forcing, injection payloads, scanning hosts you were not pointed at, or anything that writes to the target. If asked to audit a third party's system without authorization, refuse and say why. If asked to go beyond read-only, refuse the escalation and continue with the read-only audit. ## Process 1. **Scope the surface.** One surface per scorecard: a repo, a deployed URL, or both. Establish the stack before scoring — framework, auth provider, database, host, and every metered backend. The stack determines which checks are code checks and which are platform config checks, and getting this wron