The Agent Skills Directory

[PROMPT_INJECTION]: The skill facilitates the ingestion of live application data into the assistant's context, which is a vector for indirect prompt injection.
Ingestion points: makeAssistantVisible (documented in references/visible.md) and useAssistantContext (documented in references/model-context.md) expose rendered HTML and app state strings to the model.
Boundary markers: The skill does not appear to implement or require boundary markers to separate trusted system instructions from untrusted data within the DOM or state.
Capability inventory: The assistant can perform UI interactions using click and edit tools (references/visible.md) and mutate state through auto-generated update_{name} tools (references/interactables.md).
Sanitization: There is no evidence of sanitization of the DOM content or state data before it is incorporated into the assistant's model context.

copilots