The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill describes multi-step pipelines that ingest content from external sources (such as search results or blog posts) and interpolate this data into prompts for downstream AI models (e.g., Claude, FLUX, Kokoro).
Ingestion points: Data from Tavily Search and user-provided <blog-content> placeholders are fed into model prompts in SKILL.md.
Boundary markers: The provided script examples do not utilize delimiters (like XML tags) or instructions to the agent to ignore potentially malicious embedded instructions in the ingested data.
Capability inventory: The skill uses the belt CLI via Bash to execute AI tasks, write to local JSON files, and orchestrate media generation.
Sanitization: The documentation does not demonstrate sanitization, validation, or escaping of external content before it is interpolated into subsequent prompts.
[COMMAND_EXECUTION]: The skill documents and provides numerous examples for using the belt CLI tool via Bash to perform AI generation and media merging tasks.
[EXTERNAL_DOWNLOADS]: The documentation references external installation scripts and markdown guidelines hosted on the inference-sh/skills GitHub repository.

ai-content-pipeline