The Agent Skills Directory

Data Ingestion and Processing Surface: The skill is designed to process data from external sources—specifically YAML formatted fences—into agent prompts and UI state updates. This pattern is a standard architectural approach for data-driven agents but introduces a surface for indirect prompt injection where instructions embedded in the data could potentially influence agent behavior.
Ingestion points: External data is ingested through specific code fences such as yaml-spec, yaml-edit, and yaml-patch as described in SKILL.md.
Boundary markers: The library uses distinct code fence markers (e.g., ```yaml-spec and ```yaml-edit) to delimit external content from the rest of the message.
Capability inventory: The capabilities described are focused on parsing YAML and emitting JSON Patch operations for UI rendering and state management.
Sanitization: While the documentation focuses on the parsing and transformation logic, standard practice would involve validating the resulting JSON patches against a known schema to ensure integrity.
Vendor-Owned Resources: The skill utilizes libraries from the @json-render ecosystem, which are directly related to the skill's purpose and the developer's infrastructure.

yaml