The Agent Skills Directory

[SAFE]: The skill documentation describes a legitimate UI framework for AI agents. It follows safe practices for handling sensitive inputs (such as Pydantic's SecretStr) and provides standard mechanisms for state interpolation and action chaining.\n- [INDIRECT_PROMPT_INJECTION]: The skill features a surface for indirect prompt injection via the processing of untrusted data in UI components.\n
Ingestion points: Untrusted data enters the context via the state dictionary in PrefabApp, rows in DataTable, and defaults in Form.from_model (found in SKILL.md and references/forms.md).\n
Boundary markers: The skill uses {{ }} templates to delimit state interpolation.\n
Capability inventory: The skill enables CallTool, SendMessage, and UpdateContext actions (documented in SKILL.md).\n
Sanitization: While UI rendering typically escapes content, the skill does not explicitly describe sanitization of instructions embedded within processed data.

prefab-ui