The Agent Skills Directory

[PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection because it is designed to ingest untrusted data from the repository context and documentation to generate executable HTML/SVG artifacts.
Ingestion points: SKILL_CARD.md identifies user requests, repository context, architecture notes, and documentation as the primary data sources.
Boundary markers: There are no instructions in SKILL.md to use boundary markers or to ignore instructions that might be embedded within the processed files.
Capability inventory: The skill generates self-contained .html files that include inline JavaScript for interactivity and theme persistence via localStorage, as seen in SKILL.md and the references/architecture-example.html template.
Sanitization: No sanitization or escaping mechanisms are described in SKILL.md or references/style-notes.md, which could allow malicious code or instructions from the input documents to be faithfully reproduced in the generated artifact.

html-diagram-artifact