The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-supplied external materials, including URLs, reports, PDFs, notes, and transcripts (see SKILL.md under 'Supported Inputs'). This creates an attack surface where malicious instructions embedded within these source materials could attempt to influence the agent's behavior.
Ingestion points: User-supplied materials such as URLs, PDFs, and reports documented in SKILL.md.
Boundary markers: The prompt templates in references/prompts.md (e.g., 'Research Brief' and 'Outline Architect') use placeholders like {{KNOWN_CONTEXT}} or {{CONTEXT}} to interpolate data but lack explicit instructions to the agent to disregard instructions potentially embedded within that external data.
Capability inventory: The skill coordinates agent tasks including research, outlining, and artifact generation; while it does not define specific tool implementations, it orchestrates an agent that likely possesses network and file access capabilities.
Sanitization: No specific sanitization, validation, or filtering of external input is described in the workflow instructions.
[DYNAMIC_EXECUTION]: The skill includes prompt templates in references/prompts.md (Section 6) for generating SVG and HTML code based on processed content. While this is a primary feature for presentation agents, generating renderable code from external, potentially untrusted inputs is a form of script generation that requires awareness of the data being used to populate those templates.

ppt-agent-workflow