← ClaudeAtlas

psylinks-360-securitylisted

Apply the PsyLinks 360° Security Review whenever building, editing, or reviewing software that involves logins, signups, user accounts, APIs, backend services, databases, payments, file uploads, third-party packages, or any AI-agent/MCP-tool workflow with file, shell, or network access. Trigger proactively for "build a login", "add authentication", "sign up flow", "API endpoint", "backend", "database", "connect an MCP server", "add a package/dependency", "security review", "harden this", "is this secure", "prepare for launch/production" — even if the user never says "security". This is a mandatory default for this kind of work, not an opt-in add-on.
PsylinksSecurityWeb/PsyLinks-360-Security · ★ 0 · API & Backend · score 57
Install: claude install-skill PsylinksSecurityWeb/PsyLinks-360-Security
# PsyLinks 360° Security Review © PsyLinks Security Private Limited. Licensed use only — see `LICENSE.md`. This is proprietary methodology (checklists, scanner, and report format), not public-domain material; don't present it as your own or redistribute it as a separate product. See "Licensing model" below for what's actually enforceable and why. ## Why this skill exists (read once, then apply every time) Checklists alone don't work anymore, and the data shows it plainly: - Independent audits of publicly deployed AI-generated ("vibe-coded") apps found **over 90% shipped with at least one vulnerability**, averaging seven per app, most rated critical or high. - Georgia Tech's Vibe Security Radar tracks CVEs directly caused by AI-generated code — the count has been climbing every month. - Roughly a **fifth of AI-suggested package names don't exist** — attackers register those exact names ("slopsquatting") and wait. - AI-assisted commits leak hardcoded secrets at **roughly double** the rate of human-written commits. - Security research on coding agents documents "**false success**": agents confidently reporting a task as done and secure when it isn't — the agent's own self-assessment is a genuinely unreliable signal, not just a style problem. - Separately, guard logic that "looks reasonable" (a regex, a length cap, a role check) frequently protects against the wrong attack vector — code can pass every functional test while remaining exploitable. Partial security is not securi