squirrelscan

Squirrelscan is a website QA tool that audits sites for over 300 rules across SEO, performance, security, accessibility, and agent experience. It is designed to feed fixes directly to coding agents, enabling an audit–apply–re‑audit loop without manual intervention.

The tool runs on three main fronts: locally via a CLI binary, as a skill in agents like Claude Code, Cursor, Codex, and any other LLM‑based coding agent, or in the cloud with browser rendering and AI analysis. The CLI accepts a single command, “run squirrelscan”, and streams the audit report to standard output or any pipe. When used inside an agent, a simple “run squirrelscan and fix” prompt launches the audit, applies the generated fixes, and re‑runs until the site scores 100. CI pipelines can use the binary's exit code to fail a build if the score or severity thresholds are not met.

The audit engine is incremental and memory efficient. It tracks ETag, Last‑Modified headers, and uses content hashing to resume from checkpoints. It outputs a single self‑contained HTML file that can be opened anywhere; the same report is also available in console, markdown, JSON, XML, and plain text for agents. Each finding carries an LLM‑optimised fix line so that an agent can apply the change without further interpretation.

Squirrelscan's rule set spans 22 categories. Examples include accessibility checks for ARIA labels and focus management, SEO core checks for meta tags, canonical URLs, and favicon, performance core web vitals (LCP, CLS, INP) and script loading, and security checks for CSP headers and leaked secrets. Every rule is tagged with severity and provides a concrete fix. The cloud offering adds headless‑browser rendering for JavaScript‑heavy sites, AI‑powered E‑E‑A‑T authority scoring, and technology fingerprinting of 385 components such as CMS, CDNs, analytics, and payment services. The developer can run “squirrel self update” to sync to the latest rule set.

Key features

• 300+ audit rules across 22 categories
• Incremental crawling with ETag, Last‑Modified, and checksum support
• One‑prompt integration with Claude Code, Cursor, Codex, and other LLM agents
• Multiple output formats: console, markdown, llm, json, xml, text
• Self‑contained HTML report with no server dependencies
• CI‑friendly exit codes and scoring thresholds
• Cloud upgrades: headless browser, AI analysis, technology fingerprints