testing-client-side-dom-vulnerabilitieslisted
Install: claude install-skill UnboundCompute/security-agent-skills
# Testing client-side DOM vulnerabilities: the server never sees it
A whole class of vulnerabilities never touches the server: the tainted input is read
and the dangerous action taken entirely in the browser, so server-side review and
server logs see nothing. The source is a URL fragment, a message, a piece of markup,
or a global the attacker can influence; the sink is a script execution, a corrupted
object, or a leaked cross-origin fact. Testing them means tracing taint inside the
page, from client source to client sink.
## When to use
- You are reviewing a single-page app, heavy client-side JavaScript, or a browser
extension.
- Code reads the URL, a message, storage, or the DOM and writes it into a sink.
- The server sees none of the payload, so server-side testing found nothing.
## Scope check
Test pages and extensions you own or are authorized to test. Use benign markers and
your own browser session. If you can't name the authorization, stop.
## The loop
1. **Enumerate client-side sources and sinks.** List where the page reads
attacker-influenceable input in the browser (the URL and its fragment, message
data, referrer, storage, the DOM itself) and where it writes to a dangerous sink
(HTML insertion, script or event execution, navigation, object property
assignment). A source reaching a sink without sanitization is the bug.
2. **Test DOM-based cross-site scripting.** Does any client source flow into a markup
or code sink (assigning HTML, writing a