The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes belt CLI commands via Bash to interact with various AI model endpoints for generating media and text.
[EXTERNAL_DOWNLOADS]: The skill references installation instructions for the belt CLI and recommends adding related skills from the inference-sh organization on GitHub.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its workflow of chaining outputs between different AI models.
Ingestion points: Data generated by language models (e.g., Claude) is saved to files like script.json and subsequently used as input prompts for downstream tools like kokoro-tts and omnihuman-1-5.
Boundary markers: No delimiters or instructions to ignore embedded commands are present when interpolating model-generated content into new prompts.
Capability inventory: The skill possesses the capability to execute shell commands (belt) and interact with external APIs, including social media platforms (twitter/post-tweet).
Sanitization: There is no evidence of sanitization or escaping of the content produced by LLMs before it is processed by subsequent tools in the pipeline.

ai-social-media-content