The Agent Skills Directory

[SAFE]: The skill implements a standard provider for a parent application (Narova) using documented APIs.
[DATA_EXPOSURE_&_EXFILTRATION]: No sensitive data is exfiltrated. The skill correctly instructs users to provide credentials via the OPENAI_API_KEY environment variable and explicitly warns against hardcoding secrets in configuration files or logs.
[EXTERNAL_DOWNLOADS]: No third-party packages or remote scripts are downloaded. The tool/setup.sh script verifies that only standard library modules (json, pathlib, urllib.request, wave) are used.
[COMMAND_EXECUTION]: Shell commands in documentation and tool/setup.sh are restricted to setup, verification, and standard CLI usage (e.g., narova providers add). No arbitrary command execution patterns were found.
[REMOTE_CODE_EXECUTION]: There are no patterns of remote code execution. All network activity is limited to HTTPS POST requests to api.openai.com/v1/audio/speech for speech synthesis.
[PROMPT_INJECTION]: The skill processes user-supplied instructions for voice steering but implements character length limits (4096 characters) and does not interpret these instructions as commands within its own execution context.

narova-openai