The Agent Skills Directory

[COMMAND_EXECUTION]: The file scripts/auto-iterate.js contains a command injection vulnerability. It uses child_process.execSync to execute a shell command where user-supplied prompt data is directly interpolated into the command string: node scripts/ad-producer.js generate-script "${currentPrompt}". An attacker can escape the quotes using shell metacharacters (e.g., ;, $(...), or backticks) to execute arbitrary code on the underlying system.
[COMMAND_EXECUTION]: The script scripts/ad-account-manager.js contains logic to execute external shell commands via execSync to synchronize account data. Spawning subprocesses to run scripts based on internal variables creates a potential execution surface for unauthorized commands.
[DATA_EXFILTRATION]: The skill manages and transmits sensitive API credentials, including X2C_API_KEY and GEMINI_API_KEY, which are stored in local JSON files under the credentials/ directory. These credentials, along with user-generated content, are sent to external endpoints such as a Supabase-hosted API (https://eumfmgwxwjyagsvqloac.supabase.co) and Google's Gemini API. While these are established platforms, the interaction involves the access and network transmission of authentication tokens from local storage.

ai-director