bg.
Projects / CoBrowse Audit Agent
ScreenMeet deployment tooling

CoBrowse Deployment Audit Agent

Finding the deployment issue behind a failed session.

I built a browser-driven audit that samples a customer’s live site, inspects the CoBrowse runtime and security boundaries, and converts the results into prioritized findings with paste-ready fixes.

JavaScriptBrowser automationCSPPrivacyTechnical consulting

Problem and role

Most CoBrowse escalations come from integration gaps rather than a product defect: a missing bootstrap script, unregistered deployment key, blocked network directive, incomplete multi-domain setup, inaccessible iframe, or unmasked sensitive field.

I turned that troubleshooting model into a repeatable audit. It selects representative page templates, runs consistent read-only probes, separates browser evidence from settings that require console verification, and reports exactly what the customer needs to change.

What the audit checks

  • Runtime: bootstrap and core scripts, deployment key registration, version, hotkeys, and customer entry points.
  • Session continuity: multi-domain carrier placement, configured origin behavior, and same-origin iframe coverage.
  • Security policy: script, HTTPS, WebSocket, image, and frame access required by the session.
  • Privacy: explicit masks, custom mask selectors, and sensitive non-password fields that could reach the agent view.
From site sample to customer fix
  1. ScopeChoose page templates
  2. ProbeInspect live runtime
  3. ClassifyRank session impact
  4. ReportGive exact changes

Coverage is template-based: a healthy homepage does not prove that account, checkout, iframe, or subdomain experiences are configured correctly.

Evidence before assumptions

The primary probe reads the registered runtime state rather than searching page source for a key that may live in bundled JavaScript or a tag manager. A second probe listens for real Content Security Policy violations while testing the script, HTTPS endpoint, WebSocket path, and carrier frame.

The report stays honest about what the browser cannot prove. Console-only settings such as allowed domains, multi-domain mode, and canvas streaming are listed as verification items rather than reported as facts.

Report and next steps

Findings are ranked by session impact: blockers prevent connection, high-severity gaps break an active session or expose sensitive data, medium issues create degraded or fragile behavior, and low issues cover implementation hygiene.

Each finding names the affected pages, explains what the agent or customer experiences, and supplies the relevant script, carrier, masking attribute, or merged CSP change. Clean pages are included so the customer can see where deployment behavior differs.

Skills used

Web diagnostics: DOM and runtime inspection, asynchronous script loading, CSP violation analysis, WebSockets, iframes, storage state, and multi-domain behavior.

Security and privacy: least-invasive testing, PII masking review, precise evidence handling, and clear limits on what an audit proves.

Customer engineering: translating browser-level evidence into a prioritized remediation plan that developers and security teams can implement directly.