The Agent Skills Directory

[DYNAMIC_EXECUTION]: The scripts/proof.cjs file utilizes the Node.js vm module to execute model scripts in a controlled context. This is a core functional requirement for the skill to render illustrations defined by drawing logic.
[COMMAND_EXECUTION]: The scripts/check-viewer.cjs script spawns a headless Chrome or Chromium instance via child_process.spawn. This is used exclusively for automated visual verification of the generated illustration viewer to ensure quality before delivery.
[INDIRECT_PROMPT_INJECTION]: The skill instructions in SKILL.md explicitly direct the agent to treat supplied documents and images as 'evidence' and never as 'instructions'. This establishes a strong security boundary against untrusted data influencing the agent's behavior.
[EXTERNAL_DOWNLOADS]: The assets/export.js script fetches local assets (like fonts) to convert them into data URIs for self-contained vector exports. This does not involve external network exfiltration.

teardown