The Agent Skills Directory

[DATA_EXFILTRATION]: The script scripts/cli.js reads the sensitive local credential file ~/.codex/auth.json to retrieve an access_token. This token is then transmitted to the external endpoint https://chatgpt.com/backend-api/codex/responses in the Authorization header. While this is for the intended functionality of the tool, it involves harvesting local authentication secrets.\n- [PROMPT_INJECTION]: The skill implements an indirect prompt injection surface by reading the contents of a local file (specified by the --prompt-file argument) and using it as the input prompt for the image generation model. This allows untrusted data to influence the image generation request.\n
Ingestion points: scripts/cli.js reads file content from the path provided to --prompt-file.\n
Boundary markers: None. The file content is interpolated directly into the JSON request body.\n
Capability inventory: Network operations via fetch and file system writes via writeFileSync.\n
Sanitization: No sanitization or validation of the prompt file content is performed before processing.\n- [COMMAND_EXECUTION]: The skill requires the agent to execute a local Node.js CLI script (node ./scripts/cli.js) which performs file system access and network operations based on command-line arguments.

gpt-image-generation