Overview A type confusion issue exists in the @digitalocean/do-markdownit package. In the callout and fenceenvironment plugins, the allowedClasses and allowedEnvironments options are expected to be arrays of strings. If these options are provided as a single string, the code applies .includes directly on the string, resulting in substring matching instead of membership checks against an array.

Affected Versions All versions up to and including 1.16.1 (npm).

Impact Supplying crafted input can bypass intended allow-lists (e.g., class/environment constraints) due to substring checks, which may enable rendering of unintended classes or environments and lead to policy bypass in downstream consumers.

Mitigation Until an upstream fix is released, ensure configuration normalization before invoking the plugins: - Validate that allowedClasses and allowedEnvironments are arrays (Array.isArray(...)), converting single strings into one-element arrays when necessary. - Consider sanitizing or strictly validating user-controlled values that influence Markdown rendering.

Authentication is globally bypassed in github.com/nanobox-io/golang-nanoauth between v0.0.0-20160722212129-ac0cc4484ad4 and v0.0.0-20200131131040-063a3fb69896 if ListenAndServe is called with an empty token.

A command injection vulnerability exists in DigitalOcean Droplet Agent through 1.3.2. The troubleshooting actioner component (internal/troubleshooting/actioner/actioner.go) processes metadata from the metadata service endpoint and executes commands specified in the TroubleshootingAgent.Requesting array without adequate input validation. While the code validates that artifacts exist in the validInvestigationArtifacts map, it fails to sanitize the actual command content after the "command:" prefix. This allows an attacker who can control metadata responses to inject and execute arbitrary OS commands with root privileges. The attack is triggered by sending a TCP packet with specific sequence numbers to the SSH port, which causes the agent to fetch metadata from http://169.254.169.254/metadata/v1.json. The vulnerability affects the command execution flow in internal/troubleshooting/actioner/actioner.go (insufficient validation), internal/troubleshooting/command/exec.go (direct exec.CommandContext call), and internal/troubleshooting/command/command.go (command parsing without sanitization). This can lead to complete system compromise, data exfiltration, privilege escalation, and potential lateral movement across cloud infrastructure.