The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/kie_generate.py uses subprocess.run to execute curl for secure file uploads and open to reveal the results folder on macOS systems. The instructions also direct the agent to use the find command to locate recently downloaded media files in the user's ~/Downloads and ~/Desktop directories.
[EXTERNAL_DOWNLOADS]: The skill interacts with external APIs including api.kie.ai and kieai.redpandaai.co for image generation and storage. It also references Higgsfield Marketing Studio for video generation tasks.
[DATA_EXFILTRATION]: Local reference images and videos are uploaded to the KIE.ai infrastructure (redpandaai.co) for processing. This is a primary functional requirement for the image-to-image and video generation models used by the skill.
[PROMPT_INJECTION]: The skill processes advertisement content from external sources (Meta Ad Library, TikTok, Instagram, YouTube) which are treated as untrusted input. The SKILL.md file contains explicit safety warnings and requires human confirmation before any action that involves billing credits or publishing content.
[CREDENTIALS_SAFE]: The scripts/kie_generate.py script implements secure credential handling by loading KIE_API_KEY from local .env files and passing the secret to curl via a configuration stream rather than command-line arguments, preventing exposure in process listings.

clone-ad