Security audit
Security checks for vulnerabilities and agentic risk
This is a documentation-only Remix performance skill with some examples that require care, but I found no hidden execution, persistence, credential collection, or purpose-mismatched behavior.
Install only if you want Remix v2 SSR and performance guidance. Treat the window.ENV examples carefully: expose public values only and prefer the serializer pattern rather than copying the initial JSON.stringify snippet unchanged.
Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.
**Gotcha**: the `page` prop must be an **absolute path** (starts with `/`). Relative paths silently fail — no warning, no prefetch.
The example injects window.ENV into an inline <script> using JSON.stringify, which does not escape </script> or U+2028/U+2029. If any supposedly public env value contains attacker-controlled content, it can break out of the script context and lead to XSS; the surrounding skill context explicitly warns that this pattern is unsafe, which confirms the risk rather than mitigating it.
<Outlet />
<ScrollRestoration />
<script
dangerouslySetInnerHTML={{
__html: `window.ENV = ${JSON.stringify(data.ENV)}`,
}}
/>
Model output is used without validation or sanitization. Unvalidated output injected into downstream contexts (SQL, shell, HTML) enables injection attacks and arbitrary code execution.
import serialize from "serialize-javascript";
<script
dangerouslySetInnerHTML={{
__html: `window.ENV = ${serialize(data.ENV, { isJSON: true })}`,
}}
/>
The skill description includes a long list of broad trigger tokens such as generic framework APIs and common identifiers (headers export, window.ENV, window, useId). This can cause the skill to activate in many loosely related contexts, increasing the chance that an agent applies Remix-specific caching or server/client guidance where it is not appropriate, which can mislead downstream code generation or review decisions.
The prose at L239 states that a matching conditional request should return 204 to skip the body, but the sample implementation at L248-L249 returns 304, which is the correct HTTP status for If-None-Match revalidation. This is an active contradiction between documentation intent and code behavior, even though the code is correct.
The guidance recommends fixing locale-sensitive formatting by explicitly passing "en-US" in Intl.DateTimeFormat, which can be read as prescribing a specific language/locale choice. The file does not mention offering user locale selection or limiting this advice to a justified region-specific context.
The example sets <html lang="en"> explicitly, which is a natural-language locale choice. Under the policy for this audit, forcing a specific language/locale without user opt-in or documented justification is a policy violation.
No suspicious patterns detected.