The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user research, source material, and existing artifacts to build presentations, which creates a surface for indirect prompt injection.
Ingestion points: Workflow steps in SKILL.md (Step 1 and 2) instruct the agent to read and inspect user-supplied material and existing artifacts as the basis for the presentation.
Boundary markers: The instructions lack specific delimiters or directions to ignore potentially malicious instructions embedded within the source documents (e.g., hidden text attempting to override the agent's behavior).
Capability inventory: The skill possesses capabilities to write files and utilize platform-specific authoring tools to generate HTML, PDF, and native slide decks.
Sanitization: There are no explicit instructions for the agent to sanitize or escape external content before it is incorporated into the final artifact's text or generated scripts.
[DYNAMIC_EXECUTION]: The skill is designed to generate executable HTML and JavaScript to create interactive explanatory models and UI behaviors.
Script generation: As documented in references/interactive-html.md, the agent is tasked with implementing JavaScript logic for scenario models, process navigation, and data updates based on the user's subject.
Dynamic assembly: The logic for these interactions is dynamically constructed at runtime based on the specific variables and relationships identified in the user-provided data.

present