The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/scan.mjs heavily relies on execSync to run shell utilities like find, grep, and cat to perform its scanning logic.
[REMOTE_CODE_EXECUTION]: Multiple functions in scripts/scan.mjs are vulnerable to command injection because they interpolate data from the codebase being scanned directly into shell command strings without sanitization.
In the scanSecrets function, the script executes cat "${f}" where f is a filename returned by find. A file with a name like $(touch pwned).js would trigger arbitrary command execution.
In the scanNpmAudit function, the script uses cd "${absDir}", which is vulnerable if the directory path contains shell metacharacters.
In the scanDependencies function, the script runs npm view "${name}@${version}" using package names and versions parsed directly from a package.json file, allowing for injection via malicious package metadata.
[SAFE]: The skill recommends and integrates with well-known security tools such as Semgrep, Trivy, and Gitleaks.
[SAFE]: The use of official package registries via npm audit and npm view for dependency analysis is consistent with the skill's intended security purpose.

security-scan