Security context

What an agent needs to avoid regressing past fixes and find the next vuln in this repo.

davila7/claude-code-templates

              main @ 869640b
            

Fixes

CVEs

CRITICAL

Peak severity

0.0%

Coverage

Highlights

Command Injection: 3 prior fixes. Scrutinize any change in this area.

.github/workflows/discord-release-notification.yml: most-fixed (2 issues). Treat as high-risk during review.

4 high-severity fixes in this history; regressions here are high-impact.

Recurring patterns

The bug types that recur here, drawn from past fixes, not open vulnerabilities.

Remote Code Execution: The sandbox management server allowed remote code execution via the /api/install-agent endpoint by passing unsanitized input directly to spawn with shell option enabled. This provides attackers a direct vector to compromise the host system running the sandbox.

Command Injection: Using untrusted metadata inside shell runners (via direct string interpolation in 'run' steps or raw 'echo' statements) allowed arbitrary command execution in the runner context. Security requires using structured environment variables and jq or printf.

Cross-Site Scripting (XSS): Unsafe innerHTML assignments allowed arbitrary code execution in user browsers. Safe textContent assignments or rigorous output escaping are required to securely render dynamic components.