The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell commands to inspect the local environment for available API credentials.
Evidence: Uses env | grep -oE in SKILL.md to identify configured keys like OPENAI_API_KEY and GEMINI_API_KEY without printing their values.
[EXTERNAL_DOWNLOADS]: Installation instructions reference official SDKs for reputable AI service providers.
Evidence: References to google-genai, openai, replicate, and fal-client in provider-specific reference files.
[DATA_EXFILTRATION]: The skill initiates network requests to official endpoints of established AI services.
Evidence: Python scripts in references/fal.md, references/gemini.md, references/openai-xai.md, and references/replicate.md communicate with APIs for Google, OpenAI, xAI, fal.ai, and Replicate.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied natural language prompts to guide image and video generation.
Ingestion points: User prompts for image/video generation tasks specified in SKILL.md.
Boundary markers: Absent; the skill relies on natural language context rather than technical delimiters.
Capability inventory: Network operations using provider SDKs and local file system writes for saving output images (as seen in references/fal.md, references/gemini.md, and references/openai-xai.md).
Sanitization: Absent; the skill relies on the safety guardrails and content filtering of the destination model providers.

model-integration