The Agent Skills Directory

[SAFE]: The skill uses a dependency-free Python script (scripts/mintegral_ads.py) that relies only on the Python standard library, effectively eliminating supply chain risks from third-party packages.
[CREDENTIALS_UNSAFE]: Credentials are handled securely by reading them from environment variables (MINTEGRAL_ACCESS_KEY, MINTEGRAL_API_KEY) rather than hardcoded strings. The script includes a redact function that systematically masks sensitive keys such as 'access-key', 'token', and 'api_key' in all console output and error messages.
[EXTERNAL_DOWNLOADS]: All network operations are locked to official Mintegral domains (ss-api.mintegral.com and ss-storage-api.mintegral.com). The implementation uses a custom RejectRedirects handler to ensure that signed authentication headers are never leaked to external hosts through HTTP redirects.
[DATA_EXFILTRATION]: The skill implements strict local data protection. The script validates that any JSON payload or output file is a regular file (preventing symlink attacks), is owned by the current user, and has restricted permissions (mode 0600), ensuring sensitive campaign data remains private.
[COMMAND_EXECUTION]: To prevent accidental or malicious account changes, the skill enforces a mandatory preview workflow. Any mutation requires an explicit --execute flag and a replacement_plan_hash (SHA-256), which ensures the payload and the target object's state have not changed since the user reviewed the preview.

mintegral-ads