The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to clone a repository from an untrusted third-party source (https://github.com/dou-jiang/codex-console.git) and pull Docker images from a public registry (ghcr.io/dou-jiang/codex-console:latest). These sources are not verified or associated with the established vendor.
[COMMAND_EXECUTION]: The skill provides instructions to execute local build scripts (build.sh, build.bat) and database migration tools (alembic upgrade head). Executing scripts from a freshly cloned untrusted repository can lead to immediate system compromise.
[CREDENTIALS_UNSAFE]: The skill is designed to handle and store extremely sensitive data, including credit card numbers, CVVs, and expiry dates. While it mentions encryption, the exposure of these fields to an automated AI agent environment increases the surface area for credential harvesting.
[DATA_EXFILTRATION]: A core feature of the skill is the automated extraction and 'upload' of OpenAI access tokens and session data to external API gateways (e.g., CPA, Sub2API, New-API). This functionality effectively serves as a data exfiltration pipeline for high-value authentication credentials.
[DYNAMIC_EXECUTION]: The skill utilizes uv sync or pip install to download dependencies and uses build scripts to generate standalone executables at runtime, which are then used to manage account auth flows and solve security challenges (POW).

codex-console-automation