The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to clone code from a repository owned by an unverified GitHub user (asz798838958) that is not associated with a known trusted organization.
[COMMAND_EXECUTION]: The installation steps require running multiple commands with broad capabilities, including pip install, npm install, and playwright install, which fetch and install various binaries and libraries from external sources.
[REMOTE_CODE_EXECUTION]: The skill establishes a workflow where code is downloaded from a third-party repository and immediately executed via the Python uvicorn server, which can lead to the execution of arbitrary malicious code if the source repository is compromised.
[CREDENTIALS_UNSAFE]: The application is explicitly designed to store and manage plain-text passwords for ChatGPT accounts. It includes an API route (/accounts/export) that retrieves these sensitive credentials in an unencrypted format, increasing the risk of credential exposure.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect injection by processing data from external, untrusted sources such as temporary email providers and verification services to retrieve registration codes, which could potentially contain malicious payloads designed to influence the automation logic.

freeagentidentity-chatgpt-auto-register