The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted data from external sources and research to generate code-heavy output, creating a surface for indirect prompt injection. * Ingestion points: Information is ingested via the Read, WebFetch, and WebSearch tools from user-provided files and internet sources (references/intake-from-files.md, references/intake-from-research.md). * Boundary markers: Absent; the instructions do not specify using delimiters or prompts to ignore instructions within the source material. * Capability inventory: The agent generates an index.html file containing interactive Javascript figures based on code archetypes (assets/article-template.html, references/figure-archetypes.md). * Sanitization: Absent; there are no instructions for validating or escaping external content before its inclusion in the generated HTML and Javascript output.
[EXTERNAL_DOWNLOADS]: The article template in assets/article-template.html fetches typography from Google's official Fonts service (fonts.googleapis.com). This is a well-known service used for styling and does not represent a malicious external download.
[COMMAND_EXECUTION]: The SKILL.md file suggests a local Python command (python3 -m http.server 8000) for the user to preview the generated file. This is a common and standard development practice for local file verification.

creating-explainers