The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The script scripts/check_environment.py uses dynamic imports via importlib.import_module to verify the presence of required standard library modules. This usage is restricted to a hardcoded list of modules for diagnostic purposes and does not pose a security risk.
[PROMPT_INJECTION]: The skill documentation provides the agent with techniques for rephrasing user prompts to better navigate external API safety filters. This is framed as a creative optimization process and includes explicit instructions to respect safety boundaries and avoid bypassing protocols for disallowed content.
[CREDENTIALS_UNSAFE]: The skill manages API keys using environment variables and a local JSON configuration file. It follows security best practices by applying restricted file permissions (0o600) to the configuration file, ensuring it is only accessible by the current user.
[DATA_EXFILTRATION]: The image generation script performs network requests to user-defined API endpoints. It implements comprehensive URL validation, including checking hostnames against loopback and private IP ranges and enforcing the use of HTTPS, which effectively mitigates data exfiltration and SSRF risks.

gpt-imagegen