The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a vendor-specific CLI tool dna-author to perform deterministic rendering of JSON data into text, markdown, mermaid, or HTML formats. This is a standard operation within the vendor's toolchain.
[INDIRECT_PROMPT_INJECTION]: The skill processes external JSON data which is a potential surface for indirect prompt injection. However, the skill includes explicit instructions to the agent to render artifacts verbatim and avoid re-describing or summarizing the content ("The rendering must not become a second model"), which serves as a behavioral guardrail against the agent interpreting instructions embedded within the data.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill explicitly mandates that HTML output must not fetch external resources ("HTML fetches nothing at view time — inline the styling"), which prevents potential data leakage or tracking via remote assets.

dna-render