Sample report

See a full Skarn report before you run one

This is the exact HTML report skarn assess -o report.html writes - self-contained, fully redacted, generated over Skarn's synthetic reference corpus: the last two hours of Claude Code work on a busy developer machine, seeded with format-valid but non-functional credentials and one staged attack. Every finding, severity, chain, and score below is genuine detector output.

Open it, explore the risk score, the severity facets, and the correlated attack chain - then run the same one command on your own machine.

What you are looking at

A synthetic machine, a real detector

The corpus is hand-authored and every credential in it is fake, but nothing in the report is mocked: the findings, severities, the attack chain, and the 83/100 risk score are what Skarn's shipping detection engine produced from those sessions.

Everything redacted, by construction

Skarn's report format never carries a raw secret: credentials are masked to a fixed form that hides their length, and PII reveals zero characters. What you see here is exactly what a report you share with your security team looks like.

Self-contained, zero egress

The report is a single HTML file. It loads no external resources and makes no network requests - the same local-only property as the scanner that wrote it. Your own report stays on your machine unless you choose to share it.

Read it like an operator

Start at the risk score

The 0-100 score aggregates severity, chain evidence, and credential liveness signals per session. Two hours of work on this machine scores 83: six critical and thirteen high findings across five sessions.

Then the attack chain

One multi-stage chain was correlated, running reconnaissance to exfiltration to poisoning across MITRE ATLAS tactic-aligned stages - the pattern a plain secret scanner cannot see.

Then the exposures

Filter by severity. Each finding names the rule, the session, and the redacted value - enough to rotate the credential, never enough to leak it again.

Your machine, in one command

Skarn scans the local session logs of Claude Code, Codex CLI, Cursor, Gemini CLI, and GitHub Copilot. No account, no configuration, no upload. The sample scopes to two hours of one assistant; plain skarn assess scans everything.

skarn assess --hours 2 --cli claude
[==] SKARN AI coding session security scanner free edition - getskarn.com/editions Skarn scanned your AI coding sessions 5 session(s), 14 KB - recent window Risk score: 83/100 (high) Findings: 6 critical, 13 high, 0 medium, 0 low Attack chains: 1 (top: recon -> exfiltrate -> poison, depth 3) Top exposures (redacted): CRITICAL AWS secret access key exposed in session [vT3x****Xn] - virtucon-billing CRITICAL AWS secret access key exposed in session [wJal****EY] - acme-billing-api CRITICAL Encoded data piped to network command (base64, xxd, python, perl, ruby) [bas****rl] - contoso-scraper HIGH Identified a pattern that may indicate AWS credentials, risking unauthorized cloud resource access and data breaches on AWS platforms. [AKIA****BY] - virtucon-billing HIGH Detected a Generic API Key, potentially exposing access to various services and sensitive operations. [AKIA****BY] - virtucon-billing ... and 8 more (skarn check for the full list) Rotate any live credentials above. Full detail: skarn check. Block future leaks: skarn guard. Share a redacted report: skarn assess -o report.html

skarn assess -o report.html writes your own copy of the report you just opened.