The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze arbitrary HTML, CSS, and Javascript from external websites or local files to generate interaction prompts and articles.\n
Ingestion points: SKILL.md Workflow Step 1 (reading and searching external HTML/JS references for animation terms).\n
Boundary markers: Absent; no instructions are provided to the agent to treat external content as untrusted data.\n
Capability inventory: Screenshot and video capture, file system writing, starting a localhost server to host content, and execution of git and ffprobe commands.\n
Sanitization: Absent; no sanitization or validation of the ingested code is implemented.\n- [COMMAND_EXECUTION]: The workflow utilizes system commands including git (for staging and checking diffs) and ffprobe (for verifying MP4 files).\n- [PROMPT_INJECTION]: The instructions contain strict behavioral overrides and constraints (e.g., "Do not ship...", "is not acceptable", "Do not rename...") intended to force the agent into a specific, narrow workflow and output format.\n- [PRIVILEGE_ESCALATION]: The skill instructs the agent to bypass standard security and configuration boundaries. It explicitly directs the use of git add -f to bypass .gitignore restrictions and recommends serving local files via a localhost server to circumvent browser file:// navigation blocks for screenshotting purposes.

html-to-interaction-prompts