The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from multiple sources including DESIGN.md and user-supplied strings for product names, taglines, and CSS expressions.
Ingestion points: Inputs defined in SKILL.md (e.g., product_name, tagline, logo_mark) and the external file DESIGN.md (specified in Step 0 of the workflow).
Boundary markers: The skill uses template placeholders (e.g., {{PRODUCT_NAME}}) and explicit instructions to HTML-escape text and strictly sanitize SVG inputs.
Capability inventory: The skill requires file_write to generate index.html. It does not use exec, eval, or network operations.
Sanitization: SKILL.md mandates HTML-escaping for text tokens, a strict CSS grammar for color tokens (rejecting ;, {}, @, etc.), and allowlist-based sanitization for SVG inputs to strip scripts and event handlers.

waitlist-page