The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions specify executing a local JavaScript file located at a relative path (../codex-slides/scripts/codex-slides.mjs) to perform various tasks such as research, outlining, and rendering. This is a mechanism for cross-skill automation within the vendor's ecosystem.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it transforms external research findings into structured slide outlines. If the researched content contains malicious instructions, they could potentially influence the agent's behavior during the outline generation phase.
Ingestion points: The skill ingests untrusted data via the deep_research process, which is then passed as the researchDoc parameter to the create_outline function.
Boundary markers: There are no explicit delimiters or boundary markers defined in the instructions to isolate researched content from the agent's core instructions.
Capability inventory: The agent has the capability to create files, render visual decks, and navigate to URLs in a browser.
Sanitization: No evidence of sanitization or filtering is provided for the content gathered during the research rounds before it is interpolated into the slide creation workflow.

codex-slides-research